The confirmation of OpenAI's Strawberry reasoning model and the potential for it to lead to GPT-5 (Project Orion) is incredibly exciting. The advancements in AI, like GameNGen generating DOOM on the fly and Cursor AI making programming accessible to everyone, are just phenomenal. However, it's also important to consider the concerns and the push for regulation, especially with the new bill in California. Balancing innovation with ethical considerations is crucial. Thanks for keeping us updated on all these developments! 🙌
Me too, the only issue that worries me is that they are trained on Synthetic data and not real world data. That scares me for some reason. I will still use the F outa it though lol.
Cursor AI looks like a stripped-down version of Tabnine; Which has a completely free option and is basically a better version of Copilot. It seems to contain the same types of tech as a Visual Studio Code + Tabnine setup; but with a few more plugins, creating an all-in-one package to make a more user-friendly application.
It's amazing to think about what Strawberry from OpenAI and the rumored GPT-5 can do! The limits of what we can accomplish with machine learning are being pushed by the ongoing developments in AI.
I'm a professional retoucher by trade. Regarding the Pixel issue, the problem is not the fact that it can be done with Photoshop. But the reality is that even with Photoshop, it was hard to really get a good end result and required a set of skills that most people did not have. With the Pixel, it makes it super easy for anyone to do it. I don't have skin in the game either way, but even for me it takes a while to make something super realistic at a pixel level (heh). But the Pixel makes it so easy to do for anyone. It's rather interesting!
Yeah totally get that - to me the issue in this convo is more just because a technology is harder to do doesn't mean that you should discount the idea that everyone will eventually be able to do it etc. It's maybe more a question of *will* people do it?
I'm officially in the AI for Humans Discord right now, and it's about to get cooler and sexier with me around! 😎 The future of AI is going to be seriously amazing, and you don't want to miss out on the fun we're about to have. So hit that follow button and join the party because we're bringing some humor and next-level vibes to the AI community
Holy notches BatMan!!! How many holes are in Orion’s belt??? Aye, i wonder if they train like Bane for utility upgrades? We may have adopted the darkness but they were born and molded by it… #WhereDoTheyGetAllThoseWonderfulToys
AI is a beast of a tool to any independent creator. I stopped coding when I was in highshool, now im old ( you r old too) and thanks to chatgtp I'm starting new little fun coding projects and games all the time. For me, using ai to make JS apps and games is just so fun. AND the ai gen.. OR just talking shit to an AI personality. If you wanna be a billionaire, get to work on a learning program designed for the very young, using all these tools. Kids will be making shit allllllllllllllllllllllllll day nostop. I know i would if i had these tools back then.
I believe AI will radically change how easily it will be to distribute misinformation; but I also believe Humans will begin to learn how to identify the lies and just doubt everything they see online (which they should have been doing already.)
Yeah we agree with this 100% and that it should very much be a part of education going forward. I hope there's an AI literacy class in every middle school and high school very, very soon.
Personally, I suspect that AI might become its own solution to the fake-AI images problem. It wouldn't surprise me if someone made an app that could take in an image and give the user a probability that the image was either AI-generated or AI-altered in some way. Maybe it could even highlight where an AI-altered image was altered. Basically, an AI that says, "THIS LOOKS SHOPPED / I CAN TELL FROM SOME OF THE PIXELS AND FROM SEEING QUITE A FEW SHOPS IN MY TIME." This sort of tool could be built into any social media app so that images on social media platforms could be labelled automatically as AI-altered. That could really help with curbing disinformation. AI-gen will get better and better with time, but so could AI-gen-detection. Maybe it will become its own arms race. Maybe an even more advanced AI-detector could search the web to look for corroborating images or articles. Imagine someone shows a faked image of a crash on a highway, then your AI-detector says, "This photo is likely AI-altered. The highway in the image appears to be the I-95 and no crashes have been reported on the I-95 matching this depiction".
Yeah i think the cross referencing thing with real information could be vital but also your theory is the entire reason they're talking about watermarking AI images which I don't really have a ton of problem with. Of course, if you REALLY wanted to get around that you could be then you're back to using bigger tools than just the Pixel 9.
I'm always so surprised how long it can take for people to see Kevin on the show due to the algo -- we've been doing it for over a year - but always happy to have to people pop in
I have so many favorite youtube shows where unfortunately the hosts love to shit on AI. As an artist all my life and a synthographer since I first got my Dall-E invite, it's super refreshing to have somewhere without that hate. Great job guys!
hey thank you so much! we really appreciate this sort of comment and totally understand where you're coming from. also, thanks for joining the discord!
The crux of the argument about AI fakery boils down to barier of entry. In that respect I concede now just about anybody can generate a convincing fake audio, image, or video. But the issue is not that it is fake, vfx, photoshop, media editing in general is not a novel concept, what is new is the time and cost it takes to generate a realistic fiction has been greatly reduce by a technology that anybody willing can access. To my mind this is not a severe issue except in the minds of older generations who refuse to accept media should no longer be treated as convincing evidence of extraordinary claims. And even more concerning is power grab attempts that disguise themselves as solutions to that issue like world coin, where big tech seeks to insert themselves as authorities and arbitors of truth. Imo it is a temporary issue that only affects older generations bc younger generations growing up with this tech will come to be more discerning and skeptical just bc some media posted online appears to dipict some incredible reality. But I would tell older generations "do you believe everything you see on tv or in the movies just bc the vfx look convincing?"
"To my mind this is not a severe issue except in the minds of older generations who refuse to accept media should no longer be treated as convincing evidence of extraordinary claims." Couldn't agree more which I guess is weird and strange but overall something we're going to have to accept. Where it's goint to get even WEIRDER is eventually when people are living in weird entire worlds and they see completely different versions of reality but we'll cross that bridge when we come to it.
@@AIForHumansShow Fret not...I have heard tales that in the early days of cinema, a train bearing down on the cemera resulted in temporary reactiion in an audience that never experienced that before and they would dive to dodge that train bc of simple cimatograrphy pespective.
Funny how I was playing Cyberpunk while listening to you guys and once I turned off the game I was thinking that the real Cyberpunk is here outside the game as I had just finished listening to the automatic-coding situation and I was day-dreaming about the Cyberpunk legendary AIs (trapped in their world behind the wall) manipulating and creating code as part of their powers and I think that says a lot about the times we live in
This reminds me I need to play the DLC -- I love that game so much. I was one of the lucky ones who waiting until the patch came out to play it. Have you seen the latest insane realistic upscales on it?
@AIForHumansShow Nope I haven't seen the upscaled version! I'm waiting for a new pc to play it with path tracing and all the juicy things 🤤 Edit: Omg I think I just saw that mod, this is ridiculous!
This to me is the best ai show/podcast I've seen I will be joining discord and staying up to date with all episodes this was a well done show for my first time viewing 😂😂😂 them hi comments to every llm had me rolling on The floor laughing 😂😅
I don’t mean to be insulting, but there is no way Kevin is 28 years old… 38??? lol love the boys though lol I had to look it up so I wasn’t being a jerk, he’s 41 and looks stunning!
Worst explanation of MoE ever, sorry guys :) MoE is not slower, it's faster. That's one of the reasons for OpenAI using it. The slowness in Strawberry comes from its iterative nature.
3 sisters invited to eat strawberries in jobes house 🏠 I truly hope we get perfect alignment seeking eqaul measure above all else on par with a William tyndale so we don't have a bunch of miss alignment censorship issues to deal with in our future kids generations. We know how ugly that can get as many still struggle with lucid deterministic mature order while pushing the yoo hoo woo into movie evolutionary oline magical fun far from real life measure.
@@AIForHumansShow wow that's speaks volumes to English speaking ultimate precision instrument ✝️ xyzt tunes all atoms / Adam's lattus structure and body . We would be under Islamic rule or still in servitude if not the many who saught perfect alphabetical exodus + indo European language
@AIForHumansShow 1300-1500s pre issac newton plagerizing the eccentric fundamentalist Christian ✝️ xyzt formulism math mapping kinda goes a lot of Good men before Luther's reform & the enlightenment
40:30 Shout out to Kassem G
I've been using cursor for a few months now and it's been a wonderful addition to my workflow. Absolutely crushes brain block!
Yeah it's seems like a pretty huge unlock overall
super excited to try to dive into this over the weekend
I wonder if Strawberry will know how many Rs there are in its own name. 🤔
same here!!
I think that’s why it’s named that
@@thebigsteaks8752 The first major LLM with the ability to count the frequency of a given letter in a word! 🙌 I'm pretty stoked for it!
Has to be why they call it strawberry
The confirmation of OpenAI's Strawberry reasoning model and the potential for it to lead to GPT-5 (Project Orion) is incredibly exciting. The advancements in AI, like GameNGen generating DOOM on the fly and Cursor AI making programming accessible to everyone, are just phenomenal. However, it's also important to consider the concerns and the push for regulation, especially with the new bill in California. Balancing innovation with ethical considerations is crucial. Thanks for keeping us updated on all these developments! 🙌
hahah this is an amazing comment and I'd like to make sure that you are a human. can you tell me what your favorite meal to eat is?
🍕@@AIForHumansShow
"I would never speak directly to god!" haha- thank you for not cutting that out!
Hahaha sometimes the stuff you say you’re gonna cut is the best stuff to never cut
i struggled. per the usual. thanks. :)
we could never
You guys are so fun! This show is exactly what i've been searching for
Welcome! And stick around we do a lot of weird fun stuff!
7:56 I love it when a model nears competition 😂
immediate regret the moment it left my mouth... much like this sentence, likely.
Great intro! Thanks guys for keeping it fun.
shoutout to Beemop for his always great intro edits!!
OpenAI still haven’t released the tutor feature? When is that finally going to get released?
which one? did they have a tutor feature they were going to drop? actually can't remember that one.
@@gavinpurcell i forget what they called it but it was where chat gpt can see your ipad screen so when you were doing math it can see what you see?
@@ChrisCoelho that was in the apple presentation I think? so expect to see that in like mid 2028
subscribed
im so excited about the future of AI
we gona be able to make our own full movies and video games from a prompt one day
WE LOVE TO HEAR IT thanks for the sub!
welcome!
Thanks for another great ep fellas. Its been a rough week so this was a nice little escape/distraction for me 🙏🙏
oh man, we're always happy to be here as a distraction but hope everything is ok. thanks for always watching dipset
hey thanks for this and hope this week goes better
Excited for something new from OpenAI
so are we it's been long enough now
Me too, the only issue that worries me is that they are trained on Synthetic data and not real world data. That scares me for some reason. I will still use the F outa it though lol.
if synthetic data really is what people are saying than stuff is going to move even faster
I'm so upset that it took me this long to find your channel. You guys are absolutely hilarious and I learn so much. Love it!
Welcome aboard! Happy to have you -- we do this longer thing weekly and then other videos from time to time!
Hilarious and informative show. We always look forward to your humor and deep insight. Congrats on Patreon!
Hey thank you -- we appreciate this a ton!
Awww... you still got the feels fo' your gurl Kassem...💖💖💖 I need to find my wifeys video of ou to for vday that one year
Cursor AI looks like a stripped-down version of Tabnine; Which has a completely free option and is basically a better version of Copilot.
It seems to contain the same types of tech as a Visual Studio Code + Tabnine setup; but with a few more plugins, creating an all-in-one package to make a more user-friendly application.
Yeah so much of the best use cases of these tools is often who can put together the best UX and how the tools can work in an interesting way etc.
It's amazing to think about what Strawberry from OpenAI and the rumored GPT-5 can do! The limits of what we can accomplish with machine learning are being pushed by the ongoing developments in AI.
IT'S SUMMER IN THE GALAXY
I'm a professional retoucher by trade. Regarding the Pixel issue, the problem is not the fact that it can be done with Photoshop. But the reality is that even with Photoshop, it was hard to really get a good end result and required a set of skills that most people did not have. With the Pixel, it makes it super easy for anyone to do it. I don't have skin in the game either way, but even for me it takes a while to make something super realistic at a pixel level (heh). But the Pixel makes it so easy to do for anyone. It's rather interesting!
Yeah totally get that - to me the issue in this convo is more just because a technology is harder to do doesn't mean that you should discount the idea that everyone will eventually be able to do it etc. It's maybe more a question of *will* people do it?
DOOM runs faster rendered through a diffusion model than when I first installed it on my 386/25sx 😢
hahahaha this really and truly made us laugh
I'm officially in the AI for Humans Discord right now, and it's about to get cooler and sexier with me around! 😎 The future of AI is going to be seriously amazing, and you don't want to miss out on the fun we're about to have. So hit that follow button and join the party because we're bringing some humor and next-level vibes to the AI community
NOW THIS IS WHAT WE"RE TALKING ABOUT
(are you a bot)
the AI for Humans discord has everything...
@@gavinpurcell shameless
@@FriscoFatseas I am what I am
I can't wait untill OpenAI releases "Strawberry"! I have noticed GPT-4o has gotten smarter and faster! I like the creature shop, it looks so real!
hey thanks for watchin' all the way through -- and yeah it def feels like things are speeding up again
great show thank you
i love the sarcasm on the appropriately timed when the AI was speaking
thank YOU for watching
@@robodigitalis haha thank you
Holy notches BatMan!!! How many holes are in Orion’s belt???
Aye, i wonder if they train like Bane for utility upgrades?
We may have adopted the darkness but they were born and molded by it…
#WhereDoTheyGetAllThoseWonderfulToys
AI is a beast of a tool to any independent creator. I stopped coding when I was in highshool, now im old ( you r old too) and thanks to chatgtp I'm starting new little fun coding projects and games all the time. For me, using ai to make JS apps and games is just so fun. AND the ai gen.. OR just talking shit to an AI personality. If you wanna be a billionaire, get to work on a learning program designed for the very young, using all these tools. Kids will be making shit allllllllllllllllllllllllll day nostop. I know i would if i had these tools back then.
We def are old and yes, I can't wait to see what sort of weird stuff the next generation of kids make.
you have me at 'you r old too'
I garnered some inside intel: Sora will drop November 1st.
👀👀👀
Hey guys, still here :D Sorry, been busy the past week! Still appreciate you guys!
hey john thanks for popping in -- we appreciate you too!
you da best John!
I believe AI will radically change how easily it will be to distribute misinformation; but I also believe Humans will begin to learn how to identify the lies and just doubt everything they see online (which they should have been doing already.)
Yeah we agree with this 100% and that it should very much be a part of education going forward. I hope there's an AI literacy class in every middle school and high school very, very soon.
Or is it endless summer with no release?
please no... please
Personally, I suspect that AI might become its own solution to the fake-AI images problem.
It wouldn't surprise me if someone made an app that could take in an image and give the user a probability that the image was either AI-generated or AI-altered in some way. Maybe it could even highlight where an AI-altered image was altered.
Basically, an AI that says,
"THIS LOOKS SHOPPED / I CAN TELL FROM SOME OF THE PIXELS AND FROM SEEING QUITE A FEW SHOPS IN MY TIME."
This sort of tool could be built into any social media app so that images on social media platforms could be labelled automatically as AI-altered. That could really help with curbing disinformation.
AI-gen will get better and better with time, but so could AI-gen-detection.
Maybe it will become its own arms race. Maybe an even more advanced AI-detector could search the web to look for corroborating images or articles. Imagine someone shows a faked image of a crash on a highway, then your AI-detector says, "This photo is likely AI-altered. The highway in the image appears to be the I-95 and no crashes have been reported on the I-95 matching this depiction".
Yeah i think the cross referencing thing with real information could be vital but also your theory is the entire reason they're talking about watermarking AI images which I don't really have a ton of problem with. Of course, if you REALLY wanted to get around that you could be then you're back to using bigger tools than just the Pixel 9.
Wow, it's great to see Kevin back! For once, RUclips's algorithm got it right!
GUESS WHO'S BACK
I'm always so surprised how long it can take for people to see Kevin on the show due to the algo -- we've been doing it for over a year - but always happy to have to people pop in
I want AI to bring back Attack of the Show, review the latest games from GameNGen or suttin lol big props brothas keep crushin' it sirs
tytyty we appreciate it as always
Its okay guys, google pays for youtube premium to train its AI
lol good point
I have so many favorite youtube shows where unfortunately the hosts love to shit on AI. As an artist all my life and a synthographer since I first got my Dall-E invite, it's super refreshing to have somewhere without that hate. Great job guys!
hey thank you so much! we really appreciate this sort of comment and totally understand where you're coming from. also, thanks for joining the discord!
The crux of the argument about AI fakery boils down to barier of entry.
In that respect I concede now just about anybody can generate a convincing fake audio, image, or video.
But the issue is not that it is fake, vfx, photoshop, media editing in general is not a novel concept, what is new is the time and cost it takes to generate a realistic fiction has been greatly reduce by a technology that anybody willing can access.
To my mind this is not a severe issue except in the minds of older generations who refuse to accept media should no longer be treated as convincing evidence of extraordinary claims.
And even more concerning is power grab attempts that disguise themselves as solutions to that issue like world coin, where big tech seeks to insert themselves as authorities and arbitors of truth.
Imo it is a temporary issue that only affects older generations bc younger generations growing up with this tech will come to be more discerning and skeptical just bc some media posted online appears to dipict some incredible reality.
But I would tell older generations "do you believe everything you see on tv or in the movies just bc the vfx look convincing?"
"To my mind this is not a severe issue except in the minds of older generations who refuse to accept media should no longer be treated as convincing evidence of extraordinary claims."
Couldn't agree more which I guess is weird and strange but overall something we're going to have to accept. Where it's goint to get even WEIRDER is eventually when people are living in weird entire worlds and they see completely different versions of reality but we'll cross that bridge when we come to it.
@@AIForHumansShow
Fret not...I have heard tales that in the early days of cinema, a train bearing down on the cemera resulted in temporary reactiion in an audience that never experienced that before and they would dive to dodge that train bc of simple cimatograrphy pespective.
Strawberry Man? Not as good as… Celery Man 😏
can it hat wobble?!
nor as good as particle man (extreme nerd reference alert)
@@AIForHumansShow Everyone hates poor Triangle Man
Funny how I was playing Cyberpunk while listening to you guys and once I turned off the game I was thinking that the real Cyberpunk is here outside the game as I had just finished listening to the automatic-coding situation and I was day-dreaming about the Cyberpunk legendary AIs (trapped in their world behind the wall) manipulating and creating code as part of their powers and I think that says a lot about the times we live in
This reminds me I need to play the DLC -- I love that game so much. I was one of the lucky ones who waiting until the patch came out to play it. Have you seen the latest insane realistic upscales on it?
@AIForHumansShow Nope I haven't seen the upscaled version! I'm waiting for a new pc to play it with path tracing and all the juicy things 🤤
Edit: Omg I think I just saw that mod, this is ridiculous!
Kevin! the legend!
The 28 year-old legend!
more myth, but, i'll take it. :)
also, do not accept personal checks from Ai Gavin
prob also do not accept personal checks from Real Gavin
This to me is the best ai show/podcast I've seen I will be joining discord and staying up to date with all episodes this was a well done show for my first time viewing 😂😂😂 them hi comments to every llm had me rolling on The floor laughing 😂😅
tytyty so much -- these kind of comments really do make us happy
hi
right you are kev!
I just want A1 that actually does things I tell it
same larry, same
Wesley Snipes came back from a session with Blind AL.
He just wants to build a snowman 💀💩L style!!!
🤣
Or he snuck in on a previous segment where Kermit went Pacino.
I bet that pup has some BTS goods he plays while Papa Kev is away👀🐶🙈🙉🙊
Ok if ur not subscribed to this channel woop woop @1 monster milk big sass let's go my fave show u guys rock lol hi hi hi
tytyty hopefully the stars say good things about us
hi hi hi tysm
No, the issue is that the dog has a drug problem! Obviously!
MAJOR drug problem
sweet wes is in rehab now
guys im just starting out , as an AI enthusiast
would love ur feedback as i make similar stuff!
looks good keep it up!
I don’t mean to be insulting, but there is no way Kevin is 28 years old… 38??? lol love the boys though
lol I had to look it up so I wasn’t being a jerk, he’s 41 and looks stunning!
How old is Gavin tho
A.I. c what u did... hi
hi
hi
@@1pikapein hi
Worst explanation of MoE ever, sorry guys :) MoE is not slower, it's faster. That's one of the reasons for OpenAI using it. The slowness in Strawberry comes from its iterative nature.
Honestly when I watched this back before posting and I think I knew this. My bad (Gavin)
3 sisters invited to eat strawberries in jobes house 🏠
I truly hope we get perfect alignment seeking eqaul measure above all else on par with a William tyndale so we don't have a bunch of miss alignment censorship issues to deal with in our future kids generations.
We know how ugly that can get as many still struggle with lucid deterministic mature order while pushing the yoo hoo woo into movie evolutionary oline magical fun far from real life measure.
you got me to google William Tyndale
@@AIForHumansShow wow that's speaks volumes to English speaking ultimate precision instrument ✝️ xyzt tunes all atoms / Adam's lattus structure and body .
We would be under Islamic rule or still in servitude if not the many who saught perfect alphabetical exodus + indo European language
Inderectly speaking all our elusive prosperity kinda filters through such eccentric characters lol
@AIForHumansShow 1300-1500s pre issac newton plagerizing the eccentric fundamentalist Christian ✝️ xyzt formulism math mapping kinda goes a lot of Good men before Luther's reform & the enlightenment