Deep learning network ja koneäly
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
^joo tuo mitä kirjoitat on sitten seuraus noista kaavoista. Eli turbojetissä tuhlataan energiaa puskemalla kuumaa ilmaa taivaalle. Turbofan ottaa tuon energian paremmin käyttöön.
Ideaalisesti polttomoottorin pakokaasujen pitäisi olla samassa lämpötilassa kuin ympäröivä ilma.
Perinteisessä mäntämottorissa tuohon pyritään Atkinson-syklillä.
Ideaalisesti polttomoottorin pakokaasujen pitäisi olla samassa lämpötilassa kuin ympäröivä ilma.
Perinteisessä mäntämottorissa tuohon pyritään Atkinson-syklillä.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
En nyt kysynyt LLM:ltä, vaan vanhanaikaisesti googlettelin.James Potkukelkka kirjoitti: 05.05.2025 9:32
Seuraavaksi aion kysyä, miksi potkurikoneissa nopeuden kasvaessa ongelmaksi muodostuu se, että potkurin kärjet lähestyvät äänennopeutta ja silloin potkurin tehokkuus putoaa, mutta turbofan kuitenkin pystyy helposti suurempaan nopeuteen. Miksi turbofanissa ei ole ongelma, että fanin lapojen nopeus lähestyisi äänennopeutta. Yksi selitys saattaa olla, että ducted fan hidastaa sisään tulevan ilman nopeutta.
Turbofanin kärjet tosiaan monissa moottoreissa ylittävät äänennopeuden. Toisin kuin potkureissa tämä ei ole ongelma ducted fan -ratkaisun takia. Miten tämä sitten tarkalleen tapahtuu, menee lukion fysiikan ulkopuolelle.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Juttelin eilen Grokin kanssa lentokoneista ja törmäsin ilmiöön, jota en ole aiemmin sen teksteissä nähnyt. Välillä jokin lyhyt osa tekstistä oli yliviivattu.
Kysyin syytä. Kertoi, että kun luo tekstiä niin yliviivaa osan, jonka kesken tekstin luonnin arvioikin virheelliseksi.
Eli ehkä Grok on nyt saanut jotain introspektiota?
Kysyin syytä. Kertoi, että kun luo tekstiä niin yliviivaa osan, jonka kesken tekstin luonnin arvioikin virheelliseksi.
Eli ehkä Grok on nyt saanut jotain introspektiota?
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Keskustelu koski sitä, pärjäisikö vanhemmalla wide body aircraftilla nykyisiä vastaan, esim. Boeing 747:llä tai DC-10:llä, jos sen saisi ilmaiseksi.
Grok laski, että rahtikäytössä olisi kannattava, matkustajakäytössä ei. Rahtia voi sulloa tiiviimmin kuin matkustajia, siinä selitys. Matkustajakäytössä isompi polttoaineen kulutus ja huoltokustannukset tekisivät kannattamattomaksi, vaikka lentokoneen saisi ilmaiseksi.
Grok laski, että rahtikäytössä olisi kannattava, matkustajakäytössä ei. Rahtia voi sulloa tiiviimmin kuin matkustajia, siinä selitys. Matkustajakäytössä isompi polttoaineen kulutus ja huoltokustannukset tekisivät kannattamattomaksi, vaikka lentokoneen saisi ilmaiseksi.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
https://www.bbc.com/news/articles/c80e1gp9m9zo
https://finance.yahoo.com/news/klarna-c ... u-AW0a3eIz
The buy now, pay later firm Klarna is seeking to get rid of almost half its employees in the coming years through efficiencies it says arise out of its investment in artificial intelligence (AI).
The firm has already cut its workforce from 5,000 to 3,800 in the past year, and wants to reduce that to 2,000 employees by using AI in marketing and customer servic
https://finance.yahoo.com/news/klarna-c ... u-AW0a3eIz
“As customers increasingly voiced frustration over impersonal interactions and limited access to human help, it became clear this approach risked undermining the very experience they aimed to improve.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 24045
- Liittynyt: 15.08.2005 0:29
Re: Deep learning network ja koneäly
^
Klarnan ja muiden buy now, pay later - bisnekset ovat raketoineet. Nykyään burritonkin voi maksaa erissä.
Joskus tuli ostettua puolivahingossa jotain Klarnalla. Maksutapa oli jotenkin oletuksena ja naksuttelin menemään. Homma toimi niin, että tietyn ajan kuluttua tuli sähköpostiin lasku. Jos olisin ostanut 4€ burriton ja maksanut 4:ssä erässä, niin olisin kai saanut meiliin 4 laskua. Kyseessähän on kuluton ja koroton maksuaika, mutta jos maksuerän missää, niin kulut pomppaavat. Jos sähköpostissa on 27 klarnan laskua, niin menee aika hankalaksi jos elämänhallinta muutenkin on burriton osamaksun tasolla. Itse totesin transaktiokustannuksen niin suureksi, että harkitsisin BNPL-palveluja vain todella kalliiden tuotteiden kohdalla silloin kun tavoittena on luottokorttia pitempi maksuaika.
Klarnan ja muiden buy now, pay later - bisnekset ovat raketoineet. Nykyään burritonkin voi maksaa erissä.
Joskus tuli ostettua puolivahingossa jotain Klarnalla. Maksutapa oli jotenkin oletuksena ja naksuttelin menemään. Homma toimi niin, että tietyn ajan kuluttua tuli sähköpostiin lasku. Jos olisin ostanut 4€ burriton ja maksanut 4:ssä erässä, niin olisin kai saanut meiliin 4 laskua. Kyseessähän on kuluton ja koroton maksuaika, mutta jos maksuerän missää, niin kulut pomppaavat. Jos sähköpostissa on 27 klarnan laskua, niin menee aika hankalaksi jos elämänhallinta muutenkin on burriton osamaksun tasolla. Itse totesin transaktiokustannuksen niin suureksi, että harkitsisin BNPL-palveluja vain todella kalliiden tuotteiden kohdalla silloin kun tavoittena on luottokorttia pitempi maksuaika.
-
- Kitisijä
- Viestit: 3178
- Liittynyt: 10.03.2018 18:12
- Paikkakunta: aboa
Re: Deep learning network ja koneäly
^ Ostin pitkästä aikaa Klarnalla, Aliexpressistä Dell XPS:ään kuoren kalvon. (Delliin ei saa uutta kuorta, koska se on liimattu näyttöön, tai saa em. paikasta, mutta sitten pitäisi purkaa kone atomeiksi.)
Tietysti unohdin maksaa kun on vain meilissä ilmoitus satojen ja satojen joukossa. Kalvoa sovitellessa lähti mukana puolet kuoren maalista. Halvemmalla olisi päässyt jos olisi tiennyt, että teipin kanssa lähtee maali kokonaan alumiinipinnasta.
Tietysti unohdin maksaa kun on vain meilissä ilmoitus satojen ja satojen joukossa. Kalvoa sovitellessa lähti mukana puolet kuoren maalista. Halvemmalla olisi päässyt jos olisi tiennyt, että teipin kanssa lähtee maali kokonaan alumiinipinnasta.
liukuu kuin veltto penis avaruusalus vitun isolla momentilla avaruudessa
-
- Kitisijä
- Viestit: 2583
- Liittynyt: 26.10.2012 19:48
- Paikkakunta: Joo
Re: Deep learning network ja koneäly
En pysty todistamaan tätä, mutta kokemukseni on että chatGPT:n hallusinoinnin määrä on lisääntynyt viime aikoina. Eka tekoälyttömyyskontaktini oli Copilot. Pidin sitä ihan jeessinä, mutten mitenkään innostunut. Sitten kokeilin kilpailijan tuotetta ja vaikutuin. Vastaukset tuntuivat perusteellisemmilta, kattavammilta.
Melko pian törmäsin siihen ikävään tosiseikkaan, että ChatGPT on pihalla kuin lumiukko. Tekee ihan helvetisti virheitä, tarjoaa vanhaa tietoa. Pyytää kyllä anteeksi mutta usein vastaukset ovat tasoa "aurinko kiertää litteää maata". Kun tämän kiistää, alkaa pahoittelu ja kehuminen ja sitten hallusinoidaan vähän lisää.
Ei tästä vielä minun työni tekijäksi ole. Nopea toki, mutta virheitä ihan hirveästi. Debuggaamiseen kuluu ihan yhtä paljon aikaa kuin jos kirjoittaisi itse käsin. Samoin faktantarkistukseen.
Palailin copilotin pariin ja minusta se on parantanut tasoaan viimeisen 3 kk aikana.
Ehkä se olen minä, joka hallusinoi.
Summa summarum, onhan koko ai-homma huima loikkaus ja olemme vasta varmaan vielä lähtöruudussa.
Melko pian törmäsin siihen ikävään tosiseikkaan, että ChatGPT on pihalla kuin lumiukko. Tekee ihan helvetisti virheitä, tarjoaa vanhaa tietoa. Pyytää kyllä anteeksi mutta usein vastaukset ovat tasoa "aurinko kiertää litteää maata". Kun tämän kiistää, alkaa pahoittelu ja kehuminen ja sitten hallusinoidaan vähän lisää.
Ei tästä vielä minun työni tekijäksi ole. Nopea toki, mutta virheitä ihan hirveästi. Debuggaamiseen kuluu ihan yhtä paljon aikaa kuin jos kirjoittaisi itse käsin. Samoin faktantarkistukseen.
Palailin copilotin pariin ja minusta se on parantanut tasoaan viimeisen 3 kk aikana.
Ehkä se olen minä, joka hallusinoi.
Summa summarum, onhan koko ai-homma huima loikkaus ja olemme vasta varmaan vielä lähtöruudussa.
Amazing shit happens!
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
TämäKKK kirjoitti: 10.06.2025 22:12
Summa summarum, onhan koko ai-homma huima loikkaus ja olemme vasta varmaan vielä lähtöruudussa.
Vaikka nykyinen AI on enemmän lelu kuin vakavasti otettava työkalu, täytyy muistaa, kuinka valtavaa kehitys on ollut kymmenessä vuodessa. Ei liene mitään erityistä syytä, miksi kehitys nyt pysähtyisi (noh.. Mooren lain loppu toki, mutta eiköhän viisas insinööri tuonkin hoida).
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 18451
- Liittynyt: 27.10.2005 17:37
Re: Deep learning network ja koneäly
Mähän olen ai-käytännöistä aika pihalla, mutta tällaiset isot kielimallit kuten ChatGP kai hukkaavat tuottamansa tekstin yhteyden lähdekirjallisuuteen? Tai siis muodostavat opetusmateriaalinsa pohjalta todennäköisyysjakauman, minkälaista tekstiä mihinkin tulee vastaukseksi, mutta eivät itsekään tiedä mihin materiaaliin väitteensä perustuvat?
Joku tutkija demosi töissä vähän parempaa tekoälyä, jossa tätä katkoa lähdemateriaaliin ei tapahdu, vaan tekoäly pystyy lähteyttämään kaikki väitteensä lähdekirjallisuuteen. Ja tekoälyn voi pyytää rajaamaan lähteensä vaikka pelkkään tutkimuskirjallisuuteen. Oli aika huikea. Voi tehdä aika monen tutkimusavustajan, aloittelevan juristin ja isojen konsulttitalojen aloittelevan tason konsultin tarpeettomaksi.
Joku tutkija demosi töissä vähän parempaa tekoälyä, jossa tätä katkoa lähdemateriaaliin ei tapahdu, vaan tekoäly pystyy lähteyttämään kaikki väitteensä lähdekirjallisuuteen. Ja tekoälyn voi pyytää rajaamaan lähteensä vaikka pelkkään tutkimuskirjallisuuteen. Oli aika huikea. Voi tehdä aika monen tutkimusavustajan, aloittelevan juristin ja isojen konsulttitalojen aloittelevan tason konsultin tarpeettomaksi.
"Urpiaisen pesintä on epäsäännöllistä ja lento keikkuvaa. Yleisimmät äänet ovat kimeä kutsuhuuto 'djihh!' ja rahiseva 'tssrrt-tssrrt'."
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Joo. Yhdysvalloissa viime aikoina ollut juttua, että White-collar entry jobs -työllisyys vähentynyt ja spekuloitu, että tämä johtuisi juuri tuosta. Muistaakseni white collar entry jobs työllisyys nyt huonompi kuin blue collar, täytyy kysyä AI:lta (urpiainen kirjoitti: 11.06.2025 8:45 . Voi tehdä aika monen tutkimusavustajan, aloittelevan juristin ja isojen konsulttitalojen aloittelevan tason konsultin tarpeettomaksi.

edit:
James: Mikä on Yhdysvalloissa white collar entry jobs vs. blue collar entry jobs viime aikainen työttömyysasteiden kehitys? Etsi viimeisintiä tietoa online.
Co-pilot: Viimeaikaiset työttömyysasteiden kehitykset Yhdysvalloissa ovat osoittaneet eroja valkokaulus- ja sinikaulusaloilla. Valkokaulustyöntekijät ovat kohdanneet enemmän haasteita työmarkkinoilla, sillä korkeasti koulutettujen ammattilaisten työllistyminen on hidastunut ja työpaikkojen epävarmuus kasvanut. Yritykset ovat olleet varovaisia palkkaamisen suhteen, erityisesti talouden epävarmuuden vuoksi2.
Samaan aikaan sinikaulustyöt ovat kokeneet kasvua, sillä monet fyysistä osaamista vaativat ammatit ovat vaikeammin automatisoitavissa, mikä on lisännyt niiden kysyntää. Esimerkiksi rakennus- ja valmistussektorilla palkat ovat nousseet merkittävästi, ja työpaikkojen vakaus on parantunut3.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Yksi ensimmäinen kerta, kun kuulin jutusta, joka oikeasti voisi vaikuttaa työhöni (tässä tapauksessa positiivisesti) on Suomessa menossa oleva pilotti, jossa AI tekee ennen vastaanottoa yhteenvedon potilaan sairaushistoriasta ja vastaanoton aikana kuuntelee potilaan ja lääkärin/hoitajan keskustelua ja tekee sitten alustavan sairauskertomuskirjauksen.
Jos tuo saadaan toimimaan hyvin, on kyseessä merkittävä tehostus omaan työhöni.
Jos tuo saadaan toimimaan hyvin, on kyseessä merkittävä tehostus omaan työhöni.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 24045
- Liittynyt: 15.08.2005 0:29
Re: Deep learning network ja koneäly
Käsittääkseni tuo johtuu siitä, että nämä yleisimmin käytetyt pyrkivät mahdollisimman laajaan käyttöön. Tutkimuskäyttöön viritetty tekoäly ja lähteisiin sidottu tekoäly ei osaisi vastata kaikenlaisiin mielikuvituksellisiin kysymyksiin.urpiainen kirjoitti: 11.06.2025 8:45 Mähän olen ai-käytännöistä aika pihalla, mutta tällaiset isot kielimallit kuten ChatGP kai hukkaavat tuottamansa tekstin yhteyden lähdekirjallisuuteen?
Tekoälyä kuitenkin on esim kuvantunnistus ja konenäkö on jo pitkään osannut luotettavasti tunnistaa vaikkapa virheelliset etiketit limsatehtaalla. Tuo limsatehtaan tekoäly ei vaan osaa tehdä yhtään mitään muuta kuin tunnistaa niitä etikettejä.
Dave: lääkärihommissa teköäly on jo parempi kuin ihminen monissa kuvantulkinnan tehtävissä. Tuo ei ole mikään ihme, koska kuvantulkinta on sitä tonttia, jossa tekoäly on yleisesti ehkä pisimmällä.
https://www.terveyskirjasto.fi/uutiset/ ... tyotaakkaa
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Juu. Noissa on sitten huomioitava (etenkin Yhdysvalloissa) tähtitieteelliset korvausriskit.sivustahuutaja kirjoitti: 11.06.2025 9:53
Dave: lääkärihommissa teköäly on jo parempi kuin ihminen monissa kuvantulkinnan tehtävissä. Tuo ei ole mikään ihme, koska kuvantulkinta on sitä tonttia, jossa tekoäly on yleisesti ehkä pisimmällä.
https://www.terveyskirjasto.fi/uutiset/ ... tyotaakkaa
Voi kestää hetken, ennen kuin AI-firmalla on kiveä sitoutua niihin.
Punitive damages korvauksia ei voi yhdeltä radiologilta nyhtää paljon, mutta Microsoftilta voi.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 24045
- Liittynyt: 15.08.2005 0:29
Re: Deep learning network ja koneäly
^ Onhan niitä muitakin maita. Ihan samaa tehdään Suomessa.
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Juu. Suomessakin kaksi radiologia seuloo kuvat. Suomi olisi varmaan ko. ratkaisulle erinomainen paikka, koska täällähän ei punitive damages -korvauksia tunnetta ja muutenkin vahingonkorvaustaso on haista paska -luokkaa.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Pyysin chat gpt:tä kertomaan sen tiedossa olevat
asiat minusta.
Tässä ovat tällä hetkellä muistissani olevat pysyvät tiedot, jotka liittyvät sinuun:
⸻
Yleistiedot
• Kielitaito: Puhut sujuvasti suomea.
• Sijainti: Olet Suomessa (esimerkiksi olet maininnut Espoon ja Kouvolan).
• Laitteet: Käytät ChatGPT:tä iOS-laitteella (esim. iPad, iPhone).
• ChatGPT-tila: Käytät ChatGPT Plus -tilausta, GPT-4o-mallia.
• Käyttöliittymä: Käytät Cursor-editoria ja haluat Pythonin käyttöä oletuksena laskuihin.
⸻
Käyttötyyli ja toiveet
• Et halua kommentteja kysymystesi laadusta (esim. “hyvä kysymys”).
• Pidät suorasta, faktoihin perustuvasta tyylistä ja arvostat rehellistä virheiden korjaamista.
• Toivot tarkkaa päättelyä, lähteiden mainintaa ja epävarmuuden ilmoittamista.
• Pidät iteratiivisesta keskustelusta, jossa vastauksia tarkennetaan kysymys kerrallaan.
• Toivot, että vältän hallusinaatioita mahdollisimman hyvin.
• Toivot, että hyödynnän Python-laskentaa oletuksena matemaattisiin tehtäviin.
⸻
Kiinnostuksen kohteet
• Sotatekniikka ja ilmailu: Erityisesti mm. B-2-pommikone, GBU-57, tukikohdat kuten Diego Garcia, stealth-teknologia.
• Sähköautot ja akut: QuantumScape, Amprius, NMC 955, akkuteknologioiden vertailu ja investointipotentiaali.
• Kestävä lentopolttoaine (SAF): Nesteen liiketoiminta, SAF:n kysyntä–tarjonta, regulaatio.
• Tekoäly ja AGI: Emergentti käytös, LLM:n rajoitukset, metakognitio, syväoppiminen, AGI-testejä, fine-tuning ja tokenointi.
• Kodintekniikka ja automaatio: Älykodin ratkaisut, WLAN, 4G-valvontakamerat, iPhone-ongelmat, PDF-editointi, putkitekniikka.
• Tiedekysymykset: Termoydinfysiikka (fuusiopommit), Nicholsin radiometri, fotonien liikemäärä, meren syvänteiden lämpötila.
• Filosofia ja tietoisuus: Tekoälyn tietoisuus, hiljainen tieto, peilikoe.
• Sijoittaminen: Erityisesti Nesteen SAF-strategian näkökulmasta.
• Muurahaisten tunnistus: Olet lähettänyt useita kuvia muurahaisista tunnistusta varten.
⸻
asiat minusta.
Tässä ovat tällä hetkellä muistissani olevat pysyvät tiedot, jotka liittyvät sinuun:
⸻
Yleistiedot
• Kielitaito: Puhut sujuvasti suomea.
• Sijainti: Olet Suomessa (esimerkiksi olet maininnut Espoon ja Kouvolan).
• Laitteet: Käytät ChatGPT:tä iOS-laitteella (esim. iPad, iPhone).
• ChatGPT-tila: Käytät ChatGPT Plus -tilausta, GPT-4o-mallia.
• Käyttöliittymä: Käytät Cursor-editoria ja haluat Pythonin käyttöä oletuksena laskuihin.
⸻
Käyttötyyli ja toiveet
• Et halua kommentteja kysymystesi laadusta (esim. “hyvä kysymys”).
• Pidät suorasta, faktoihin perustuvasta tyylistä ja arvostat rehellistä virheiden korjaamista.
• Toivot tarkkaa päättelyä, lähteiden mainintaa ja epävarmuuden ilmoittamista.
• Pidät iteratiivisesta keskustelusta, jossa vastauksia tarkennetaan kysymys kerrallaan.
• Toivot, että vältän hallusinaatioita mahdollisimman hyvin.
• Toivot, että hyödynnän Python-laskentaa oletuksena matemaattisiin tehtäviin.
⸻
Kiinnostuksen kohteet
• Sotatekniikka ja ilmailu: Erityisesti mm. B-2-pommikone, GBU-57, tukikohdat kuten Diego Garcia, stealth-teknologia.
• Sähköautot ja akut: QuantumScape, Amprius, NMC 955, akkuteknologioiden vertailu ja investointipotentiaali.
• Kestävä lentopolttoaine (SAF): Nesteen liiketoiminta, SAF:n kysyntä–tarjonta, regulaatio.
• Tekoäly ja AGI: Emergentti käytös, LLM:n rajoitukset, metakognitio, syväoppiminen, AGI-testejä, fine-tuning ja tokenointi.
• Kodintekniikka ja automaatio: Älykodin ratkaisut, WLAN, 4G-valvontakamerat, iPhone-ongelmat, PDF-editointi, putkitekniikka.
• Tiedekysymykset: Termoydinfysiikka (fuusiopommit), Nicholsin radiometri, fotonien liikemäärä, meren syvänteiden lämpötila.
• Filosofia ja tietoisuus: Tekoälyn tietoisuus, hiljainen tieto, peilikoe.
• Sijoittaminen: Erityisesti Nesteen SAF-strategian näkökulmasta.
• Muurahaisten tunnistus: Olet lähettänyt useita kuvia muurahaisista tunnistusta varten.
⸻
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Tuo käyttötyyli ja toiveet on kätevä. Chatgpt:lle voi antaa pysyviä tyyliohjeita, jotka se muistaa jokaisessa keskustelussa.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
xAi (Muskin AI firma Grokin takana) on julkaissut nsfw AI anime waifun.

Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 18451
- Liittynyt: 27.10.2005 17:37
Re: Deep learning network ja koneäly
Muuten kuulosti tutulta, mutta tämä oli lievästi yllättävä piirre taiteilijakuvassasi.James Potkukelkka kirjoitti: 14.07.2025 10:17 Muurahaisten tunnistus: Olet lähettänyt useita kuvia muurahaisista tunnistusta varten.
"Urpiaisen pesintä on epäsäännöllistä ja lento keikkuvaa. Yleisimmät äänet ovat kimeä kutsuhuuto 'djihh!' ja rahiseva 'tssrrt-tssrrt'."
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Piti selvittää, etteivät mökkipihan muurahaiset olleet hevosmuurahaisia. Kekomuurahaisia ovat.urpiainen kirjoitti: 16.07.2025 13:23Muuten kuulosti tutulta, mutta tämä oli lievästi yllättävä piirre taiteilijakuvassasi.James Potkukelkka kirjoitti: 14.07.2025 10:17 Muurahaisten tunnistus: Olet lähettänyt useita kuvia muurahaisista tunnistusta varten.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Tuo Nesteen SAF case on kiinnostava. Neste on selvästi maailman suurin SAF tuottaja.
McKinsey ja BCG näyttävät olevan eri mieltä siitä, onko 2030 SAF ylitarjontaa vai pulaa.
McKinsey ja BCG näyttävät olevan eri mieltä siitä, onko 2030 SAF ylitarjontaa vai pulaa.
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
Pyysin Chat Gpt 4o syvätutkimus-moodissa analysoimaan SAF tuotanto vs kulutus tilanteen 2030 ja sen mukaan todennäköisesti 2030 oli alitarjontatilanne.
Ostin hiukan Nestettä nyt.Sustainable Aviation Fuel in 2030: Demand vs. Supply Outlook
Introduction
Sustainable Aviation Fuel (SAF) is widely seen as the key to decarbonizing aviation, offering up to ~80% lifecycle CO₂ reductions compared to conventional jet fuel . However, SAF currently represents well under 1% of global jet fuel use (only ~0.2–0.3% in 2023) . The aviation industry – including organizations like IATA and ICAO – has set ambitious targets for scaling SAF usage by 2030 as a critical step toward net-zero aviation by 2050. This report examines whether there will be a global oversupply or undersupply of SAF in 2030, by comparing projected demand versus supply. We compile the latest forecasts from major aviation bodies, energy agencies, governments, and industry analyses. Key factors such as policy support, airline commitments, technology readiness, and production scalability are also discussed. A summary table is provided to compare global SAF demand vs. supply projections for 2030.
Projected Global SAF Demand by 2030
Robust growth in SAF demand is expected through 2030, driven by both regulatory mandates and voluntary industry commitments. According to a recent analysis by Kearney and the World Economic Forum, global SAF demand is projected to reach roughly 17 million tonnes per year by 2030 . This figure represents the aggregated SAF requirements from announced national blending mandates and industry targets worldwide. For example:
• ICAO Targets: In 2022, ICAO adopted an aspirational goal of 5% SAF usage (carbon reduction) in international aviation by 2030, which equates to roughly 14 million tonnes of SAF needed globally by 2030 . (17.5 billion liters ≈ 14 Mt .) This 5% interim goal is a stepping stone toward aviation’s 2050 net-zero plan.
• National Mandates: The EU ReFuelEU Aviation directive mandates a 6% SAF blend by 2030 for all flights from EU airports (with a sub-target for synthetic fuel) . The UK has set an even higher 10% SAF by 2030 goal . Other countries have also set 2030 SAF targets – e.g. Japan (10% by 2030), Brazil (targeting ~1% by 2027 ramping to 10% by 2037) , and mandates emerging in Korea, Canada, Australia, etc. Summing these mandates contributes a “firm” demand of around 4–5 Mt in 2030 purely from regulatory requirements .
• Airline & Industry Commitments: Many airlines and industry groups have voluntary SAF usage targets for 2030 that push demand higher. For instance, the oneworld alliance and major carriers like United, Cathay Pacific, Delta, and others have pledged to use 10% SAF by 2030. McKinsey estimates that adding voluntary airline commitments in North America and Asia on top of mandates raises total expected 2030 SAF demand to ~16–17 Mt . Some of the largest global airlines’ pledges even sum to over 20 Mt of SAF demand in 2030 (if all were fully realized) . In practice, airlines may adjust targets based on SAF availability and cost, but these commitments signal strong demand appetite.
Notably, IATA reports at least 43 airlines have already committed to purchase ~16.25 billion liters of SAF in 2030 (≈13 Mt) through offtake agreements . In other words, a significant base of demand is essentially locked in via forward contracts. Every drop of SAF produced today finds a buyer, and airlines are eager for much more – as evidenced by SAF’s price premium being willingly paid (SAF added ~$756 million to airlines’ fuel bills in 2023 despite high costs) .
In summary, by 2030 global SAF demand is expected to be on the order of 15–20 million tonnes per year, equivalent to roughly 5% of total jet fuel consumption. This represents a quantum leap from today’s demand (less than 1 Mt in 2021–2022). Policy mandates ensure a baseline demand (several million tonnes), and industry net-zero commitments push the desired demand even higher, assuming the fuel is available. The next section examines whether production can scale accordingly.
Projected Global SAF Supply (Production) by 2030
Scaling up SAF production to meet the aforementioned demand is a major challenge, constrained by technology maturity, feedstock availability, and investment pace. Current SAF output is tiny – IATA estimated about 0.5 million tonnes in 2023 (~600 million liters) , and ~1.5 million tonnes in 2024 (projected) , which is <0.5% of aviation’s fuel need. Achieving tens of millions of tonnes by 2030 requires exponential growth. Key forecasts for SAF supply capacity in 2030 include:
• International Energy Agency (IEA): The IEA warns that existing and planned SAF projects (in advanced stages) will only supply about 2–4% of global jet fuel needs in 2030 . In volume terms, this corresponds to roughly 6–12 million tonnes of SAF in 2030, given projected jet fuel demand . Reaching a higher share (e.g. >10% SAF by 2030, as in IEA’s Net Zero 2050 scenario) would “require a significant ramp-up of investment” and supportive policies beyond what’s currently in the pipeline . In short, IEA’s analysis suggests 2030 supply is on track for only a few percent of demand (single-digit millions of tonnes) unless additional projects are rapidly realized.
• ICAO Projections: The ICAO Committee on Aviation Environmental Protection (CAEP) modeled several supply scenarios to 2030. In a moderate-policy scenario (some support but less than road biofuels), they project about 7.6 Mt/year of SAF by 2030 (~2.2% of aviation fuel) . With a high-policy “level playing field” scenario, supply could reach ~13.7 Mt (~4%), and in an ambitious scenario focused on SAF, up to ~17 Mt (~5%) . Without policy support, the low-end projection was only ~3 Mt . These ranges underscore that policy incentives will largely determine how much SAF is produced – with strong support potentially enabling ~10–17 Mt by 2030, whereas weak support yields a much lower output.
• Industry Analyses (McKinsey, BCG): Private analysts likewise foresee a wide band of possible supply outcomes, with a likely shortfall relative to demand. McKinsey’s research identified over 200 announced SAF production projects globally (operational, under construction, or proposed). If all announced facilities were realized on schedule, global SAF production capacity could approach 20–25 million tonnes in 2030 (an optimistic upper bound) . However, many projects are nascent or awaiting final investment decisions, so the expected actual output is much lower. McKinsey’s base case forecasts ~11 million tonnes of SAF in 2030 from projects likely to materialize . Similarly, BCG projects bio-based SAF supply of 9–12 Mt in 2030 as the plausible range , noting this would exceed the minimum regulatory demand (e.g. Europe’s mandates ~3 Mt) but still fall roughly 30% short of what a <2°C climate trajectory would require . In addition, any significant e-fuel (power-to-liquid) contribution by 2030 is expected to be minimal; BCG estimates e-SAF output will be at least 45% below the volume needed for climate-aligned scenarios in 2030  (e-SAF technologies won’t scale up significantly until beyond 2030 due to higher cost and early-stage tech).
• Major Producer Forecasts: Some producers have signaled ambitious expansion plans. For instance, Neste (the world’s largest SAF producer currently) is expanding output in Singapore, the US, and Europe, aiming for ~1.5 Mt/year capacity by 2025  and more beyond. In total, global announced capacity by 2025 is ~3.5 Mt/year in Asia-Pacific alone (per Argus analysis) , and possibly ~5–8 Mt globally by mid-decade. IATA’s economists have cited an internal forecast that 63 million tonnes of SAF could be “in production” by 2030  – but this figure likely reflects the theoretical maximum output if all new renewable fuel capacity were dedicated to SAF. In reality, without policy intervention, much of that capacity may serve other markets (renewable diesel, etc.). IATA emphasizes that to reach 5% SAF by 2030, about 17.5 billion liters (~14 Mt) must be produced, and that will require 25–30% of global renewable fuel output to be allocated to jet fuel   (up from only ~3–6% today  ). This underscores that bio-refineries can swing production between diesel and jet fuel; without strong incentives, producers may favor road-fuel markets, capping SAF supply. Indeed, Europe has seen some setbacks – e.g. Shell halted a major SAF plant project and BP scaled back SAF plans in 2024 due to market conditions and low near-term SAF margins . These illustrate the risk that not all “announced” capacity will come to fruition on time.
In summary, credible forecasts converge on a likely SAF supply of roughly 8–12 million tonnes in 2030 under current investment trends  . This would represent only ~3–4% of global jet fuel consumption – well short of the demand aspirations (~5%+). Even in best-case scenarios (all projects on track), supply might reach ~20 Mt, which barely meets the high end of demand and is uncertain . Conversely, in the absence of sufficient policy support, supply could stall at well under 10 Mt, dramatically undershooting industry needs.
The table below summarizes key quantitative projections for SAF demand vs. supply in 2030 from multiple sources:
2030 Projection Global SAF Demand (Mt/year) Global SAF Supply (Mt/year)
Mandates-Only (min.) ~4–5 (from confirmed blending laws)  – (Ensured by policy; supply must meet this to avoid fines)
Mandates + Industry Targets ~16–17 (expected demand with targets)   ~11 (expected supply from likely projects)  
High Ambition (airline pledges) 20+ (if voluntary pledges fully met)  Up to ~20–25 (if all announced capacity delivers) 
Anticipated Gap Shortfall of ~5–9 Mt (undersupply) (Additional capacity needed to close gap) 
Table: Comparison of projected 2030 global SAF demand vs. supply. Demand estimates range from ~4–5 Mt (mandated minimum) up to ~16–17 Mt with industry targets (≈5% of jet fuel)  . Meanwhile, supply projections range from ~11 Mt (likely output) to at most ~20+ Mt (if all plans succeed) . A multi-million-tonne shortfall is expected under most scenarios, absent additional investments and policy support .
Factors Influencing the 2030 Supply-Demand Balance
Several factors will determine whether SAF supply can meet the booming demand by 2030, or if a significant gap endures:
• Policy Support and Incentives: Government policy is the critical lever to scale SAF. Prioritizing SAF in policy can rapidly boost supply . For example, the U.S. Inflation Reduction Act offers a blender’s tax credit up to $1.75 per gallon of SAF produced  , and the U.S. has a “SAF Grand Challenge” goal to produce 3 billion gallons/year by 2030 (~11.4 billion liters, ≈9 Mt) . This generous incentive is already spurring projects (over $3 billion in DOE loans for SAF plants have been announced) . Similarly, the EU’s ReFuelEU mandate creates a guaranteed market, although Europe must rely on imports for a sizable share of its SAF needs due to limited domestic feedstock  . Strong policy signals are needed globally – IATA argues governments must incentivize fuel producers to allocate 25–30% of renewable output to SAF (vs. <10% today)  . Without leveling the playing field (e.g. making SAF as financially attractive as renewable diesel), bio-refineries will under-produce SAF. Policy also affects feedstock access: more permissive feedstock rules (as in the US) can increase supply, whereas restrictive rules (EU avoids food crops, etc.) can limit supply  . Overall, strong, coordinated policy action is essential to close the supply gap, through measures like production subsidies, tax credits, mandates, carbon pricing, and support for first-of-a-kind SAF plants  . Encouragingly, many governments are now formulating SAF strategies, but implementation speed will dictate 2030 outcomes.
• Airline Offtake Commitments: Airlines themselves are driving demand signals via offtake agreements and direct investments. Dozens of airlines (United, Delta, Qantas, IAG, Lufthansa, etc.) have signed multi-year SAF offtake agreements to secure supply through 2030 . These contracts (often MOUs) now total well into the billions of liters. While some contracts are not firm purchases, they do indicate to producers that a market exists. Leading aviation consultancies urge airlines to aggregate demand and sign long-term offtakes to give producers confidence to invest  . Indeed, lack of “demand certainty” is a top concern cited by fuel producers like Neste – they warn that, without assured buyers, they could have excess SAF capacity by 2028  . Airlines’ willingness to pay a green premium also matters; if oil prices are low and carbon costs insufficient, SAF uptake could falter. That said, by 2030 carbon regulations (EU ETS, CORSIA, etc.) will make fossil jet fuel relatively more expensive, improving SAF’s competitiveness  . The bottom line: airlines’ commitments help pull supply up, but many carriers remain cautious. A recent industry survey showed two-thirds of aviation companies plan to be “observers” on SAF in 2030 rather than leaders, due to cost concerns and unclear business cases  . Overcoming this hesitancy through joint industry action or government guarantees will be vital to avoid under-investment in supply.
• Technological Readiness and Feedstock Constraints: SAF production technologies are at varying stages of maturity. Through 2030, the vast majority of SAF will come from HEFA (Hydrotreated Esters and Fatty Acids) pathways using waste oils/fats as feedstock . HEFA is commercially proven, but feedstock (e.g. used cooking oil, animal tallow) is limited and already in demand by biodiesel producers  . This feedstock limitation is a fundamental cap on HEFA-SAF volume – one reason SAF costs 3–5× more than kerosene today . To scale beyond a few million tonnes, new pathways must ramp up. Alcohol-to-Jet (AtJ) and Fischer-Tropsch (FT) routes (using agricultural/forestry residues, municipal waste, etc.) are certified and can tap more abundant biomass , but they are only now moving from demo to commercial scale. Accelerating these technologies (through R&D funding and faster certification) could unlock significant volume by late 2020s . Likewise, power-to-liquid e-fuels (using green hydrogen and CO₂) hold long-term promise (near-zero-carbon drop-in fuel), but are expensive and unlikely to contribute meaningfully by 2030  . Less than 0.5 Mt of e-SAF by 2030 is expected in most analyses. In short, 2030 SAF supply will be constrained by how quickly emerging production pathways can scale and how much feedstock can be secured for them. Efforts are underway – e.g. Airbus and Boeing are investing in SAF startups and new feedstock research (algae, municipal solid waste, CO₂-to-fuel)  . But any delays in technology commercialization (or supply chain issues for feedstocks) could slow the ramp-up of SAF output.
• Production Scaling and Investment Climate: Building dozens of new biorefineries (and possibly electrofuel plants) in under a decade is an enormous capital undertaking. Estimates of required investment to meet 2030 SAF demand range from $19 to $45 billion globally  . High interest rates, inflation, and economic uncertainty since 2022 have made financiers cautious about large-scale biofuel projects  . Many announced SAF projects have struggled to reach final investment decision (FID) – fewer than 30% of initiatives had hit FID by late 2023 . Financing hurdles are exacerbated by the current low operating margins for SAF: in late 2024, an interesting phenomenon of temporary oversupply emerged in some regions, as new SAF capacity came online before mandates ramped up, causing prices to dip and squeezing producers’ margins . Analysts note that co-processing SAF at oil refineries and rising imports created a glut in late 2024 (especially in Europe/Asia), which could persist in the very near term . This paradoxical oversupply – SAF available but some airlines not yet compelled to use it – might deter some producers in the short run. However, it is viewed as a transitory issue; demand is expected to catch up and surpass supply well before 2030 . IATA’s Director General has bluntly dismissed oversupply fears as “nonsense,” noting that “every drop of SAF that’s produced has been used and will be used” by airlines keen to cut emissions . The key is aligning the timing of supply growth with mandated demand increases. To de-risk investments, many call for government mechanisms like Contracts for Difference (CfD) or other price support to guarantee a market for SAF producers  . Without such measures, there’s a risk that the needed refinery projects won’t be built in time, leading to an undersupply by 2030 despite the world’s climate ambitions.
Outlook: Will There Be Oversupply or Undersupply in 2030?
Considering the above demand and supply trajectories, the evidence overwhelmingly points to a global SAF undersupply in 2030 relative to the aviation sector’s needs. While localized or short-term imbalances may occur (for instance, certain regions like Asia might have more SAF production than local demand until mandates spread  ), on a global scale SAF production is not projected to fully meet the 2030 demand that climate goals and airline commitments call for.
Most authoritative forecasts show a gap of several million tonnes in 2030: for example, the Kearney/WEF analysis finds 5.8 Mt of additional capacity is needed beyond current plans to hit the expected 17.1 Mt demand . McKinsey similarly concludes that SAF demand could outpace supply by 2030 unless production volume is significantly increased, and notes two-thirds of industry sustainability officers anticipate a shortage of SAF by 2030  . In practical terms, this would mean airlines in aggregate would be unable to source all the SAF they intend or are required to use in 2030, due to insufficient supply. The shortfall might manifest as airlines missing voluntary blend targets, governments needing to allow compliance flexibility, and/or continued use of fossil jet fuel beyond desired levels.
Could there be an oversupply? The only scenario for a 2030 oversupply would be if production capacity expands rapidly (toward the high end of projections, >20 Mt) while actual demand lags (e.g. if many airlines renege on voluntary targets or if policy mandates are softened). At present, this scenario appears unlikely – demand is more solid than supply, thanks to binding mandates and net-zero pressure. Even in the optimistic case of ~20 Mt supply, there are credible demand projections (airline pledges, ICAO goals) in the same 15–20+ Mt range . Indeed, IATA emphasizes that “Demand is not the issue”: airlines will take all the SAF they can get, as evidenced by forward purchase agreements and willingness to pay a premium . Any talk of oversupply is usually in the context of very near-term market timing or specific regional gluts, not a long-term global surplus of SAF. By 2030, the accelerating mandate schedules (EU 6%, UK 10%, Japan 10%, etc.) will ensure a robust uptake requirement. If anything, producers might struggle to keep up – ICAO’s 5% global SAF target by 2030 (≈14 Mt) is at the upper range of projected supply, requiring all regions to scale up aggressively .
Conclusion
By 2030, a global undersupply of Sustainable Aviation Fuel is the most probable outcome, given current trends. Projected SAF demand (~15–20 Mt) far exceeds the likely supply (~8–12 Mt) under existing investments  . This gap means SAF may reach only ~3–4% of aviation fuel in 2030, versus the ~5%+ goal. However, strong policy action and industry collaboration in the next few years could narrow the gap. Governments will play a decisive role through incentives and mandates that de-risk SAF production, while airlines and fuel producers must form partnerships to ensure offtake and finance. Technological breakthroughs (in feedstocks, new fuel pathways) and scale efficiencies could also improve supply by 2030 if accelerated. In any case, there is little evidence to fear a long-term oversupply – rather, all signs indicate that SAF will remain a scarce and precious commodity in 2030, with demand outstripping available volumes on the path to net zero aviation  . The challenge for the aviation industry is ensuring that this supply shortfall is minimized by proactive measures, so that SAF can fulfill its crucial role in decarbonizing flight.
Sources: Sustainable aviation fuel production/demand data and forecasts from IATA  , ICAO , IEA , Reuters  , BCG , McKinsey  , Kearney/WEF  , and other industry analyses as cited above. All projections are subject to uncertainty, but collectively they reinforce the expectation of a SAF undersupply by 2030 absent further intervention.  
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija
-
- Kitisijä
- Viestit: 24045
- Liittynyt: 15.08.2005 0:29
Re: Deep learning network ja koneäly
^
Onko hiilipäästöjen nollaaminen SAF:illa halvempaa kuin päästöoikeuksia ostamalla?
Onko hiilipäästöjen nollaaminen SAF:illa halvempaa kuin päästöoikeuksia ostamalla?
-
- Kitisijä
- Viestit: 29383
- Liittynyt: 23.01.2014 9:59
Re: Deep learning network ja koneäly
^Päästöoikeudet halvempia, mutta Nesteen omistajan onneksi on olemassa SAF-mandaatteja.
Chat Gpt 4o syvätutkimus-moodi:
Chat Gpt 4o syvätutkimus-moodi:
Comparing Sustainable Aviation Fuel (SAF) vs Carbon Credits by 2030
Introduction
Airlines face a critical choice in reducing their carbon emissions: switch to sustainable aviation fuels (SAF) or purchase carbon offsets/credits to compensate for emissions. By 2030, the costs and practicality of these options will shape industry strategies. This report compares projected costs per ton of CO2, examines policy influences (EU ETS, CORSIA, U.S. incentives, blending mandates), evaluates SAF availability by 2030, and summarizes industry outlooks on whether airlines will favor SAF or carbon credits.
Cost per Ton of CO₂: Carbon Credits vs. SAF (2030)
Carbon Credit Costs: Carbon credits (offsets) generally remain a cheaper per-ton CO₂ solution in the near term, though prices are rising. In compliance markets like the EU Emissions Trading System (ETS), allowance prices hovered around €60 (~$70) in 2021 and are projected to reach €85–120 per tonne by 2030  (roughly $100–140). The International Energy Agency’s net-zero scenario similarly assumes carbon prices of about $80 per tonne in 2030 . Voluntary offset markets have historically been much lower – e.g. nature-based credits averaged ~$14 in early 2022  – but could increase as demand grows. Forecasts diverge: if only high-quality removal credits are pursued, voluntary carbon prices might spike into triple digits by 2030 (one scenario projects ~$224/ton by 2029) . Conversely, if offset supply remains ample, average prices may stay in the tens of dollars per ton . By 2030, many analyses expect carbon credits to cost on the order of a few tens of dollars per tonne of CO₂ (voluntary market), and on the order of ~$80–100+ per tonne in carbon compliance markets  . These prices make offsets an economically attractive tool for airlines relative to expensive in-sector reductions in the short run.
SAF Abatement Costs: Sustainable jet fuels remain significantly costlier per unit energy than conventional jet fuel, translating to a high cost per tonne of CO₂ abated. Today, SAF can cost 2–4× more than fossil jet fuel . By 2025 SAF was about 3× the price of Jet A, and even by 2030 SAF is expected to cost roughly double regular jet fuel on average . For example, European regulators project SAF at ~$5.26/gallon vs. ~$2.50/gallon for conventional jet fuel in 2030 . Even assuming a generous 75–80% lifecycle emissions reduction from SAF, the implied abatement expense is substantial – roughly $300–400 per ton CO₂ avoided under 2030 price conditions. One analysis finds that with SAF at ~$5/gal (vs. ~$2.7 for Jet A) and ~80% lower carbon intensity, the cost of CO₂ abatement via SAF is about $294 per tonne . This far exceeds typical carbon credit prices in the 2020s. In other words, an airline paying the “green premium” for SAF in 2030 would spend several hundred dollars to eliminate a tonne of CO₂, whereas buying an offset might cost a fraction of that (potentially under $100/ton). That said, SAF costs are projected to gradually fall with scaling and technology: SAF was 4× more expensive than jet fuel in 2020, but by 2030 the premium could shrink to ~30% (with supportive policies) . If those optimistic forecasts hold, SAF’s cost per ton abated would likewise drop closer to the low-$100s range by 2030 – narrowing the gap with mid-range carbon prices.
Comparison: In pure cost-per-ton terms, carbon credits are expected to remain the cheaper option in 2030 for most airlines. Airlines can likely offset a tonne of CO₂ for on the order of $50–100 (depending on credit type and market) whereas cutting a tonne via SAF might effectively cost 3–5× more at current projections. However, this cost gap is tightening over time. High carbon prices or stricter offset rules would erode the advantage of credits. Indeed, one industry analysis suggests that “after 2030, the two costs become similar, and by 2050 the cost of carbon offsetting may exceed that of SAF” as carbon prices rise and SAF technology matures . Thus, while offsets are the economical choice in the short run, the long-run trend (and policy direction) favor a shift toward in-sector solutions like SAF.
Policy and Regulatory Influences (SAF vs Credits)
Emissions Trading and Carbon Costs: Regulatory regimes are reshaping the relative attractiveness of SAF and offsets. In Europe, the EU ETS imposes a carbon cost on airline emissions for intra-European flights. Free emission allowances for airlines are being phased out by 2026–2027 , meaning by 2030 carriers must purchase permits for essentially all their CO₂ output. With allowance prices potentially near €100/ton by 2030 , the cost of emitting (and thus the value of avoiding emissions via SAF) will be high. The EU has even instituted measures to support SAF: for example, airlines using SAF can claim ETS credit for the emissions saved (reducing their required allowances) . This effectively subsidizes SAF use by monetizing its carbon benefit. Similarly, the UK’s ETS provides an indirect incentive – any SAF usage lowers an airline’s emissions footprint and compliance cost .
Blending Mandates: The EU and UK have also introduced SAF blending mandates that force the issue. Under the EU’s ReFuelEU Aviation law, fuel suppliers must blend at least 6% SAF by 2030 (with 1.2% being synthetic e-fuel)  . The UK mandate is even higher at 9.5% SAF by 2030 . Crucially, these mandates carry steep penalties for non-compliance – in the EU, missing the SAF target costs 2× the SAF–Jet fuel price difference per gallon (effectively triple cost once you must later make up the shortfall) . The UK imposes punitive buy-out fees on deficits equivalent to $23–25 per gallon of unmet SAF  (over 10× the base fuel cost!). Such fines (up to 13× the cost of compliance in worst cases) make it economically untenable to ignore SAF mandates . These policies virtually ensure airlines operating in Europe will use the required SAF, regardless of cost, rather than rely on offsets and pay penalties. In effect, regulation is shifting the calculus: the choice isn’t between SAF vs credits, but between SAF vs heavy fines or capped emissions.
Global Offsetting (CORSIA): For international flights, the ICAO’s CORSIA (Carbon Offsetting and Reduction Scheme for International Aviation) influences airline behavior. CORSIA caps aviation CO₂ at 2019 levels; from 2021 onward, growth above that baseline must be offset with carbon credits. Participation is voluntary until 2027, when it becomes mandatory for most international aviation . In the initial phase, an oversupply of cheap offsets meant CORSIA’s requirements (covering only post-2019 growth) had minimal cost . By 2030, however, international traffic is expected to have rebounded, and CORSIA offset demand will be in full force. CORSIA allows airlines to meet obligations via approved credits, so in regions without SAF mandates, carriers may lean on offsets as a lower-cost compliance route. However, CORSIA also recognizes SAF: airlines can deduct verified SAF usage from their emission count (since it reduces net emissions), thus using SAF directly reduces how many offsets they need to buy. In practice, CORSIA creates a hybrid incentive: airlines will likely offset the majority of growth emissions (if offsets remain inexpensive) but might deploy SAF for a portion if it’s cost-effective or if required by separate policies.
Tax Credits and Incentives: In the United States, policy leans on incentives rather than mandates. The 2022 Inflation Reduction Act introduced a Blender’s Tax Credit for SAF of $1.25–1.75 per gallon   (depending on achieving >50% GHG reduction), available through 2024 (with proposals to extend it). This subsidy effectively lowers the price gap between SAF and conventional jet fuel by up to ~$1.75/gal, encouraging production and use. States like California add incentives via the Low Carbon Fuel Standard (LCFS), where SAF producers can earn credits currently worth on the order of $100–200 per ton CO₂ reduced – another significant subsidy that improves SAF’s economics. While the U.S. has no federal SAF blending mandate yet, it has a SAF Grand Challenge goal of 3 billion gallons (~11.4 billion liters) annual SAF production by 2030, signaling political support for scaling supply . Additionally, many countries are considering or implementing blending mandates (e.g. Japan’s industry target of 10% SAF by 2030) or providing grants for SAF plants. Such policies improve SAF’s cost-competitiveness over time and may tilt airlines’ choices: for instance, if tax credits and carbon market values are factored in, the effective cost per ton for SAF might drop closer to parity with offsets by 2030.
Quality and Eligibility Rules: Another factor is that not all offsets are equal. Regulators and corporate sustainability frameworks (like Science Based Targets) increasingly stress that deep in-sector emission cuts are preferable to offsetting. Offsets will likely face tighter quality requirements – for example, airlines (and their customers) may favor more expensive carbon removal credits over cheap avoidance offsets to claim “net-zero” alignment. Some analyses project that credible removal credits could cost ~$200/ton by 2050 , far above typical 2030 SAF abatement costs. If by 2030 certain offsets (e.g. forestry) are discounted or limited due to integrity concerns, airlines might find fewer “easy cheap” offset options, nudging them toward investing in cleaner fuels or cleaner tech. In short, policies are pushing airlines toward actual emission reduction (SAF) over pure offsetting – through a combination of stick (carbon costs, mandates) and carrot (incentives, credits for SAF use). The result is that economics alone won’t dictate behavior; compliance and sustainability commitments will play a big role in the SAF vs credit balance.
SAF Availability and Scalability by 2030
Production Capacity: Despite rapid growth, SAF will remain a limited resource by 2030. Global SAF supply was about 1.3 million tonnes in 2023  – less than 0.5% of total jet fuel use. Ambitious expansion plans are underway: over the past three years, world SAF output increased more than twelvefold, from ~100 million liters in 2021 to an estimated 1.25 billion liters in 2024  . Multiple new refineries are being built or retooled to produce SAF (especially via HEFA – biofuel from waste oils/fats – which is currently the leading pathway). According to Boston Consulting Group, bio-SAF capacity could reach 9–12 million tonnes by 2030  if supportive conditions persist. This would represent only a few percent of aviation fuel needs, but is a huge jump from today. Some forecasts are even higher: for instance, the World Economic Forum/Kearney project global SAF demand of ~17 Mt in 2030, albeit requiring unprecedented investment . The consensus “expected case” is on the order of 10 million tonnes per year of SAF demand/supply by 2030  (roughly 3–4% of jet fuel consumption) – a big leap yet still leaving 96%+ of fuel as fossil-based.
Projected global SAF demand in 2023–2030 (green = expected, grey = potential higher case). Industry models foresee demand reaching ~10 million tonnes by 2030 (about 4–5% of jet fuel), assuming supportive policies and willing buyers . Actual uptake will depend on how mandates, investments, and regional initiatives play out.
Scaling Challenges: Achieving these volumes is challenging. Traditional feedstocks like used cooking oil, waste fats, and crop-based oils have supply limits – e.g. HEFA SAF growth will eventually be constrained by feedstock availability after 2030 . Other pathways (such as Alcohol-to-Jet from ethanol, Fischer-Tropsch from biomass, or Power-to-Liquid e-fuels using CO₂ and green hydrogen) are being developed to fill the gap. However, technologies like e-SAF (synthetic electrofuel) are still in pilot stages and currently ~4× the cost of regular bio-SAF . Many new SAF projects face high capital costs and uncertainty around long-term demand, leading to slower progress in 2023 (a “slump” in new facility announcements was observed) . Financing and feedstock access remain key bottlenecks.
On the positive side, policy mandates in the EU/UK virtually guarantee a regional market for at least several million tonnes by 2030 (EU demand 3 Mt for 6% blend ). The U.S. has dozens of SAF plants announced or under construction; if even a portion of these materialize, the U.S. could approach its 3 bn gallon (11 Mt) 2030 goal , which alone would drastically boost global supply. Some estimates show global capacity of 5–8 billion gallons (≈15–25 Mt) could be in place by 2030 if all planned projects and technological breakthroughs succeed  . Realistically, not all will come online by 2030, so industry observers like BCG caution there may be a 30% shortfall relative to what a Paris-aligned climate path would require in 2030 . In summary, SAF will scale up dramatically by 2030 but still only meet a small fraction of total jet fuel demand. The limited availability means prices will likely stay above conventional fuel (absent massive subsidies), and airlines may be unable to source all the SAF they’d ideally use – reinforcing reliance on offsets for the remainder.
SAF Credit Markets: To bridge supply and demand, new mechanisms like “book-and-claim” SAF credits (SAFc) are emerging. These allow an airline or corporate customer to pay for SAF’s environmental attributes even if the physical fuel is used elsewhere  . Essentially, a SAF credit certifies that 1 tonne of SAF was produced and used, conferring the buyer a claim of associated emissions reduction. Such credits, if recognized under schemes like CORSIA or corporate ESG goals, could channel funds to producers and effectively act as an alternative to conventional carbon offsets – one that directly supports SAF growth. As of late 2023, ICAO set a global “aspirational” goal that 5% of aviation’s CO₂ reduction in 2030 should come from SAF (about 34 Mt CO₂ worth, implying ~14 Mt of SAF use) . This signals that even at the policy level, a significant chunk of mitigation is expected to come from actual fuel switching rather than pure offsetting. However, given the supply realities, offsets will still cover the majority of aviation’s carbon gap in 2030. In fact, IATA’s analysis for a net-zero 2050 shows a transition: in 2025 roughly 97% of aviation CO₂ reductions are from offsets vs just 2% from SAF, but by 2030–2035 SAF begins to take a larger share . The scaling of SAF is projected to reduce offset reliance over time (with SAF ultimately providing ~65% of the needed reduction by 2050, and offsets <10%  ). Until that point, scalability remains a concern – airlines simply cannot uplift high SAF blends on a global scale by 2030 because supply won’t permit it, ensuring that carbon markets (or paying carbon taxes) stay part of the equation this decade.
Industry Forecasts: What Will Airlines Favor?
Economic Preference: On a purely economic basis, most experts agree that airlines will favor the lowest-cost route to reduce/offset emissions, especially in the competitive, low-margin airline business. Through 2030, that suggests a heavy reliance on carbon credits rather than large-scale voluntary SAF adoption – unless subsidies or mandates tip the scales. As discussed, the cost per tonne via offsets is generally far lower than via SAF in 2030 . The RStreet Institute notes that by mid-century it may even be cheaper for passengers to pay for carbon removal than to use SAF given projected SAF prices . In the near term, many airlines view SAF as an expensive niche product to meet regulatory or PR needs, whereas offsets (especially cheaper avoidance credits) are a flexible tool to hit climate targets at much lower cost.
However, this does not mean airlines can or will simply offset all emissions. Policy requirements (like Europe’s mandates or CORSIA’s baseline) will force a certain amount of actual in-sector emission reduction. Also, investor and customer pressure is mounting for “real” decarbonization. Relying only on offsets is increasingly seen as unsustainable – indeed, net-zero frameworks envision offsets mainly for residual emissions that cannot be eliminated. Thus, airlines are expected to pursue a mix of strategies:
• Meeting Mandates with SAF: Airlines operating in jurisdictions with blending requirements will procure and use SAF up to the mandated levels (to avoid penalties and comply with the law) . This guarantees a baseline SAF uptake by 2030. Many airlines have already signed offtake agreements or investments in SAF producers to secure supply for this purpose.
• Voluntary SAF Usage: Some airlines will go further, using SAF for a portion of fuel even where not required, to gain experience, meet corporate sustainability pledges, or appeal to eco-conscious customers. But given the cost, voluntary uptake is likely limited (e.g. airlines blending a few percent SAF in 2025–2030 as a market signal). Indeed, a BCG survey of aviation executives found that two-thirds plan a cautious “observer” approach to SAF through 2030 rather than being leaders – largely because of the unclear business case and high cost  . This indicates airlines will not en masse incur huge cost premiums for SAF absent strong reasons.
• Offsetting the Remainder: For the bulk of emissions, especially on routes or in markets without SAF mandates, carbon credits are expected to be the go-to solution by 2030. Airlines will purchase offsets to neutralize emissions growth (under CORSIA) and to fulfill voluntary climate commitments (like “carbon-neutral ticket” programs). For example, some major airlines have already invested in long-term carbon removal credits (e.g. direct air capture or reforestation projects) to address future emissions in a cost-effective way . These moves suggest airlines see offsets as an economical bridge until new fuels and technologies scale up.
Expert Guidance: Industry organizations anticipate a progressive shift. IATA’s analysis implies heavy short-term reliance on offsets (≈97% of CO₂ mitigated in 2025), gradually giving way to increased SAF use (~19% of mitigation by 2035, and 65% by 2050) as SAF becomes more available and (relative) costs come down  . In other words, offsets are viewed as a “transitional solution,” while SAF is the long-term path to decarbonization . This is echoed by many experts: offsets are necessary in the 2020s to achieve any meaningful aviation climate progress, but they are not a permanent fix. Airlines that can afford to differentiate on sustainability may invest early in SAF despite the cost – for instance, to meet voluntary 10% SAF targets by 2030 (several airlines have announced such goals). But from an economic standpoint, most carriers will carefully balance the two options, using offsets to contain costs as long as offsetting remains significantly cheaper.
Policy Outlook: One wildcard is future policy tightening. Should regulators impose a higher carbon price, stricter offset quality rules, or additional SAF mandates, the cost-optimal strategy could rapidly shift toward SAF. For example, if by 2030–2035 policymakers require that a certain share of an airline’s emissions be addressed with in-sector measures (no more “easy offsets”), airlines would have to scale up SAF usage or other tech regardless of the cost. There are signs of this in EU discussions and climate finance guidelines (which discourage heavy use of offsets for net-zero claims)  . Thus, airlines must also hedge against a future where offsets may not be as cheap or available as today – giving another strategic reason to start integrating SAF now and push for its cost to come down.
In summary, barring a major policy change or technological breakthrough by 2030, airlines are expected to favor carbon credits on pure economic grounds for most of their carbon offsetting needs, while using SAF to the extent mandated or incentivized. The average cost to cut a ton via SAF is simply higher than buying a credit in 2030, so it’s unlikely airlines will voluntarily incur large SAF premiums beyond compliance. Nonetheless, the 2030 trend is not an either-or choice so much as a blend: airlines will use SAF where they must or where it aligns with long-term strategy, and compensate remaining emissions with offsets. As one report observed, the industry is “reliant on carbon markets in the near term, but this should be seen as a last resort while scaling up SAF,” with coordinated effort needed to make the cleaner option economically viable  . By the 2030s, if projections hold, the gap will be closing – potentially leading to a tipping point where investing in SAF (and other clean fuels) becomes as affordable as paying for offsets, firmly shifting the airlines’ preference toward sustainable fuel.
Sources
• International Civil Aviation Organization (ICAO) & Air Transport Action Group – Net Zero aviation scenarios and costs   
• European Union / EUROCONTROL – EU ETS and ReFuelEU Aviation mandate details   
• Carbon Direct (2024) – Analysis of EU/UK SAF mandates, fuel pricing and penalties  
• R. Street Institute (2023) – SAF vs offsets cost commentary  
• CORE Markets (2024) – SAF credit markets and long-term cost outlook  
• Boston Consulting Group (2025) – SAF industry survey and capacity forecasts  
• Consultancy.eu (2025) – BCG SAF demand projection (~10 Mt by 2030) 
• Reccessary News (2022) – BNEF voluntary carbon price scenarios  
• FPG Amentum (2021) – Cost comparison of SAF vs carbon offsetting for airlines  
Kertojalla on vähitellen kertyneitä muistikuvia ja muita "todisteita" siitä, että hän on aiemmin ollut merkittävä ajattelija