Sam Altman New statement On GPT-5 Is Surprising!

Join My Private Community - / theaigrid
🐤 Follow Me on Twitter / theaigrid
🌐 Checkout My website - theaigrid.com/
Links From Todays Video:
• Microsoft Build 2024: ...
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) [e-mail]
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[@DicksonPau96]

You should at least name who the guy (Kevin Scott) is lol

[@TheAiGrid]

Yes i forgot to add this

[@TheAiGrid]

CTO Of microsoft

[@peterxingularity]

😂

[@francescoambrosino1841]

​@@TheAiGridDo you think that in 2035 with the advent of AGI (Artificial General Intelligence) and ASI (Super Artificial Intelligence) they will introduce universal basic income!? a subsidy!? there will be free will!? an era of abundance will begin!? the intellectual and motor capabilities of robots have already surpassed the average human and before 2035 there is talk of them surpassing those of any human and then again those of any human combined! regarding both body and mind! answer me please...

[@frustasistumbleguys4900]

The bald guy

[@SanderFocus]

Let me summarise Sam Altman's statements: 1. GPT-5 will be smarter 2. Don't miss out 3. It's amazing

[@winstonsmith6204]

13 minutes to explain this

[@MrNobodyX3]

Never base opinions off of graphs that have no labels

[@sunlight8299]

😂 noted

[@w00dyblack]

lol

[@ThisGuyDude]

And don't even show where the y-axis goes relative to the end of the x-axis.

[@elivegba8186]

thank apple for these kinda graphs

[@joecaves6235]

Never confuse a graphic with a graph. That is a graphic of money printers level of brrrrr

[@jefferybenzos5879]

Someone put "just another INSANELY CRAZY day in AI" on a shirt

[@clayferguson4776]

All these AI KZbinrs only know two adjectives: "Crazy" and "Insane". Those to words are becoming the new "like"....a word that exists in pretty much every sentence and like nobody like notices and nobody like knows how to like stop saying it.

[@patrickohara227]

@@clayferguson4776pandering to the masses; language most people use for when something excites them. More views. It is annoying. I’ll make the shirts.

[@graylienz]

Ask an ai to create a business plan creating and selling those exact shirts. It would be an interesting experiment.

[@sahulianhooligan7046]

​@@clayferguson4776dont forget Stun and shocks

[@patrickohara227]

@@graylienz I’m wary of your interesting experiments…

[@user-lm4nk1zk9y]

According to the graph, we *potentially* might be still alive in 2026. Wow

[@kamikeserpentail3778]

I plan to be alive into 3026

[@phen-themoogle7651]

@@kamikeserpentail3778 If you could you would live 1000 years?

[@SirCreepyPastaBlack]

​@@phen-themoogle7651depends on if i have freedom and a reasonable potential for happiness

[@pablopastrana4285]

the graph says nothing like that lol

[@brianjanssens8020]

​@@phen-themoogle7651i just wanna live until i'm tired of living and not be restricted to a certain age limit.

[@robweaver8872]

The chart is NOT referring to intelligence over time, it is referring to compute power required to run new models. Which is not a good thing necessarily... The chart is showing how much more computing power is required for these models as they get bigger. Again, it is NOT based on how much smarter they are getting.

[@phen-themoogle7651]

True, it doesn't corelate evenly since optimization is important. But compute power can make them bigger and lead to them becoming smarter either through optimization or through sheer size. Scaling up in tokens seemed to teach them new abilities for different types of data, and the bigger they are scaled up the more they have inside them (compare how quality of AI video generation improved) . Definitely it can be disorganized and they can hallucinate from it, but imagine how much smarter they will be if they are both hundreds of trillions of parameters and even just half optimized, like a human brain is hundred of trillions of neurons (maybe more) , right now GPT-4o claims it is 1.7 trillion parameters (old gpt4 is optimized at 175 billion , so the GPT-4o is only 10x bigger). I think something will click when it's 100x bigger than what we have now, even if just 10x bigger doesn't make it appear much smarter (still noticeable though). There might be a point where they unlock new types of intelligence or a non-human conscious/sentience that no species has had before. Maybe even above our version of sentience at some point. Compute is required for reaching new limits though, whether or not it's a good thing will be determined by what happens in a couple years , around Microsoft's 100 billion dollar stargate project and seeing what happens to AI if/when it reaches AGI and ASI. Hopefully it's smarter in a helpful way to society.

[@JordanMetroidManiac]

But there is a positive correlation between compute and intelligence

[@qaesarx]

But wait, not only is it correlated, but a smarter model WILL lower computing power because AI find new optimisations. Its a feedback loop! Now GPT4o ALREADY is 10x lower per cost! This will only lower by efficiency and data amount/inference ratio.

[@gunsarrus7836]

That chart was compute needed to train the models not run them

[@UltraK420]

@@phen-themoogle7651 Agreed. It takes foresight to understand what's happening here, but the correlations are also just obvious. This isn't like upgrading from a Windows XP computer to a Windows 11 computer, it's more like upgrading from ANY computer to a computer that hosts highly advanced digital intelligence with potential to become sentient. There's really no comparison there, AGI/ASI is a totally unconventional domain. These computers won't just compute; they think.

[@w00dyblack]

Look, I still need to go to work on Monday - tell me when this shit replaces that and I can chill instead

[@seanrobinson6407]

My thoughts exactly.

[@mrnoblemonkey8401]

You might only have to wait a year or two. But if they are wrong about the exponential growth it will be a very very long time.

[@JA-gz6cj]

You're gonna chill in the street corners in no time

[@jackzhang387]

depends on wha u do, it might take s long time if u r a nurse

[@yungouda]

“The only problem with the chart” is that it does have any scale or labels on the Y axis lol

[@phen-themoogle7651]

Sam Altman say anything new? He says that in every interview "It will be smarter across the board"....yeah? surprising? no... but I didn't watch every minute because I just heard him say the exact same thing...why are you surprised??

[@joecaves6235]

Sam is like if the middle school class soccer enthusiast, who admired the way Steve jobs stood next to epic engineers demanding they do the impossible, while taking their credit.

[@MaryjaneAmaral]

Hey Superman??? Can u help me???

[@MaryjaneAmaral]

I need a genius , cuz I'm not tech savvy, and someone is using their pixel phone to steal my data onto hers, I pay for my Motorola , etc ..android ....she uses , I think Apache, but can't we all? I need her to stop...I'm sick of the obsessed drooling desperate , sociopath , bothering me!!! How is this always allowed, I get close , then it all changes, it's a pixel though, and funny it's a phone I was gonna buy, but I don't want anything they have, cuz they listen n watch me. Illegally ...and had to if heard and seen me into pixel, cuz she had an iPhone prior, unless ,it's a new stalker! Please , simplify , I just wanna change profile pic on my KZbin, k , she changed it, and now as I try to again , change it back, it won't accept photo gallery size ??? Weird, never happened before...I'm tired, but if u know how I can add a photo gallery pic into my KZbin , as I've done before, but why now it's not allowing it due to size??? Sorry , just know you guys are naturals, hope ur not an opp, against me , if so ignore this msg...thank u

[@OmicronChannel]

2:07 In general, increasing the compute exponentially means that, close to the end of Moore’s Law, the cost also tends to increase exponentially. We should also keep in mind that inference scale basically linear with model size (according to the scaling law paper of OpenAI).

[@TheMrCougarful]

It gets interesting when the amount of compute required to train a next generation model drops by 90%. When that happens, it points to ASI. That's what I'm watching for, little else matters.

[@alexanderbrown-dg3sy]

A more efficient set of parameters eh? Would be as huge a breakthrough as backprop itself I would guess..but according to the OG Yann Lecun…that isn’t possible . Guess we shall see 😂. The same way a LM gives axial capabilities from next word prediction…I think what you speak of would be the same thing. A transformer based Gödel machine. Your statement would be a trait of such an architecture.

[@IsaacAndHisGhost]

Accept UAP'S. They matter 👽🛸🔫

[@TheMrCougarful]

@@alexanderbrown-dg3sy If Lecun did say that, then he has either gone over to the dark side, or his head is in a bucket. We have NO IDEA what is possible, or where this might end up. I'm not sure ASI is even an issue, it might solve all our problems, my point is that it will come at us out of left field. It will be here and over with before we even hear the sirens.

[@theterminaldave]

I wonder what comes after ASI? I expect companies will want to put that level of compute to something?

[@palmtreeleebythesea]

@@theterminaldave what is ASI?

[@mrd6869]

This will be near AGI...which means the work applications will ALSO jump as well. Robotics included. I believe people are gonna underestimate this thing and get caught slippin. White collar jobs have a target on their back.

[@ivandelossantos5056]

Blue collar jobs jobs will then be next, once this is integrated with robotics.

[@TheMrCougarful]

That sounds about right.

[@Ed-ym4tu]

People ARE underestimating. Vast majority do not understand where we are at, much less where we are going and definitely not how fast we will be there.

[@davidvickers8425]

​@CdawgAMVsFilmEditingoblivious is fine.

[@flashraylaser157]

​@@ivandelossantos5056I can't wait. Take all the jobs, please; then the world's governments will have no choice but to issue UBI (universal basic income) in their place so that the economy continues to function. I think someday we'll laugh that we all feared losing the jobs when it was the best and most freeing thing to ever happen to us, the earliest signs of a post scarcity world in which all we have to do is learn, travel, have fun, and perhaps produce art. All while AI takes care of everything we were once enslaved to.

[@Qui6Below]

I hate the fact that we’re using underpowered versions of these LLMs i swear we’ll never get consumer level agi

[@graylienz]

I think we will, but only time will tell.

[@CeresOutpost]

Right - We haven't even seen the half of it yet. Altman is constantly regulating his words in interviews because he knows so much more than the rest of us.

[@phen-themoogle7651]

@@CeresOutpost Idk, he said gpt-4o is "magical" and it's barely better than what we had.

[@alexhein6386]

Agi won’t happen until 2030s

[@daffi.mp3]

@@phen-themoogle7651 this is probably the first rollout, then an upgrade after the election. GPT 5 in December or Spring 2025, my predictions tho

[@soundproductionandadvice]

In my plus acc I created a gpt, loaded a pdf for it to reference, added custom cmds forcing a PDF check before responding - it still performs badly.

[@patrickohara227]

Ive noticed the same.

[@CG64Mushro0m]

as eleizer yudkowsky said in his ted talk, at some point, these companies racing ever faster to scale ai will cough out something smarter than humanity

[@TheMrCougarful]

It might be helpful to NOT think of it in human terms at all. At the end of this race, they cough up an alien intelligence that flirts like ScarJo.

[@JohnSmith762A11B]

@@TheMrCougarfulNow we talkin’!

[@nesseihtgnay9419]

GPT 4o is already massively insane, just imagine GPT 5 and beyond.

[@alexhein6386]

Those are just chat models

[@Sinoxqq]

​@@alexhein6386except it's not just chatting anymore, and a self feeding loop is also not just a chatbot

[@shamicentertainment1262]

@@Sinoxqqis 4o self feeding?

[@BionicAnimations]

@@Sinoxqq Exactly. The thing is talking now like a human, and it's about to perform tasks and beyond. Some people just don't get it.🙄

[@hobbypsychologist6444]

I hope gpt5 will be able to do actual tasks on you pc or smartphone

[@BornToTroll-it5ju]

the chat gpt 8 announcement will just be a cyborg nobody can distinguish from Sam that he's sent to give the talk while he sips martinis in St.Tropez.

[@deutschblitz]

thats awesome haha. :)

[@Macorelppa]

2:21 Man that Image is edited ❌

[@therainman7777]

That’s not a real quantitative chart. It’s just conveying the general concept that it’s going up. In fact it’s not even increasing exponentially in the chart; the beginning looks exponential but then it quickly becomes linear. Kind of like a softplus function.

[@Kuatro444]

Sam gives me that "villain" vibes, its eerie man

[@juanvaladez5703]

Listen to your intuition. He is more evil than you can imagine.

[@shadi6484]

these "Sam gives villain vibes" were at first funny, but now they just...aren't anymore. like, what has he even done to give you a villain vibe? he actually has this innocent vibe whenever he listens to someone. he also looks even younger than his age, which, again, makes him look childish and innocent. maybe it's all acting, who knows. but at least he gives off a really nice, intelligent and heroic vibe that makes you want to trust him

[@juanvaladez5703]

@@shadi6484 You know who else gives a friendly vibe that makes you want to trust him? Satan, that’s who.

[@hoffmaestro88]

⁠​⁠@CdawgAMVsFilmEditing This comment might age but I feel like we ought to give him a shot for being in the right one for this job and surpass the vile judgement based on physicality because it’s simply idiosyncratic. Remember when he got sacked by the board, 800 hundred members of OpenAI threatens to mass resign if Sam will not be back where he supposed to be. At least that’s a great sign of trust the best of us has an effort to give. What he can envision through his eyes might be something we need more than anything else, comparing to many other people out there today

[@shadi6484]

@@juanvaladez5703 dude, by this logic, there is no true friend in the world cuz they all are Satan. Like what kind of a logic was that. this was the most childish response I'd seen in a while. I'm starting to think that people like yourself who criticize Sam way too much don't have the slightest idea what you're talking about. you have no idea about his character nor his visions, and you probably haven't even watched his many interviews to actually get to know him a little bit. if you had, you wouldn't have called him "evil." Sam is an ambitious person, which isn't necessarily a bad thing. it was thanks to people like him that the world is so advanced now. and I'm sure Sam is one of the fewest people in this whole damn world whom we can entrust AGI with. He knows damn well what he's doing. he's smart, he loves AI for the sake of AI not for the sake of money and he certainly isn't evil lmao some of you guys watch too much cartoon i guess. real life is different from cartoons, please grow up. Sam is leading the world towards AGI whether you call him a villain or not. so, maybe it's better to stop wasting your life and you, too, do something big for this world, eh?

[@octavioavila6548]

I’m skeptical. I think we are in fact very close to diminishing returns with AI

[@Wandering_Bear]

There is surprisingly little information here, it boils down to Altman's opening phrase, LLMs will get smarter quickly.

[@irollerblade13]

But will Ai correct feedback noise? Actually nvm wont matter I lost all ability to hear high frequencies due to listening with headphones @1:25

[@area51xi]

Maybe should have asked GPT “hey does my audio sound ok?”

[@kas8131]

I didn't see the surprising part

[@Mandy633-se9ky]

No limitations, no boundaries, seize the dat right???,

[@alexanderbrown-dg3sy]

About 100x more compute for gpt5. Regardless of what they say at a press show multimodal…currently has limitations. Obviously 100x more compute will get meaningful gains, this is the caveman way to scaling performance. But current research shows multimodal models scale differently than pure text models. OpenAI has showed no evidence they solved for hierarchal structure..which I believe to be the solution(for the most part)…so obviously gpt5 will be dope but I don’t believe it will be a gpt-3 to gpt-4 magnitude jump. Caveman solutions have limitations. I also expect them, since they lack imagination to simply package gpt5 with “Microsoft” vasa-1 model. That ish is easy asf to train.

[@pjtren1588]

Bottle rocket, Saturn v. Both only rockets.

[@ogr3dblade6]

Not really. They were struggling for compute for a long time which is why GPT-4 was limited. They're not just upping the compute and hoping for the best, it's a physical requirement for them to continue developing the LLM.

[@kelvinndeti5482]

I really appreciate your work by the way Keep it up brother. Your information is accurate and reliable, you don't have click baits and you do your research very well. Keep it up and also thankyou we need more like you

[@samlongman1831]

Please can someone help me... I can't find an answer anywhere... I'm in the UK using android chatgpt 4o and have a premium account. I can't seem to real-timeme conversion or real-time visual input like in the recent demos. Is this ability only available to some? All I can do is voice input, and it reads back the response. This isn't the same as the demos. It's still really just a text conversion. Thank you in advance. It's driving me nuts trying to find an answer to this!

[@chrism6904]

It has not been released yet. Hopefully soon.

[@andersonsystem2]

Computer power is an important factor in the development of AI, but it's just one piece of the puzzle when it comes to creating truly intelligent machines. It's a complex and multi-faceted challenge that requires insights from a wide range of fields, from computer science to psychology to philosophy. But with continued research and innovation, I'm confident that we'll continue to make progress towards that goal, and I'm excited to see what the future holds! So hopefully we will see AGI.

[@ekurisona663]

we haven't had a platform shift in a while..... looks back at the last 125 years 💀

[@AVADAMS1967]

Grammar police -- "There ARE no limits"

[@gabrielgaidos7015]

Shouldn't that obviate your grammar?

[@elsavelaz]

What is clear a developer using the api, is the way it’s enabled full stack to where even vector db and ml capabilities are abstracted and cost absorbed in large part by OpenAI , is HUGE

[@ekurisona663]

I was literally using 4o 2 days ago to try to run a d&d campaign and it couldn't even track the number of goblins or the number of arrows I had used😂

[@Herp0]

I personally have not had this issue? But also GPT4o is not that much better then turbo...this ain't gpt5. I believe 4o is really just supposed to be multimodal and easier to run *real-time" once open AI gets past that " coming in a few weeks hurdle lol

[@ogr3dblade6]

Same. My GPT doesnt even log memories correctly. It lies that it has but it never appears in the log when its supposed to. Also it deletes its entire memory tree when it feels like it :P All that and we have to pay to use it! only to get 80 messages per 3 hours. What a rip off, defo cancelling subscription

[@JustMe-ty2rp]

The version of 4o currently available is lacking many features and not the 'big' 4o, afaik. That being said, things like hallucination and memory in AI are definitely still an issue; not sure when that will be solved, but I don't think it will be too long given how many resources are being thrown at it.

[@ogr3dblade6]

@@JustMe-ty2rp the 4-o is just to shut us up while they refine 5

[@JustMe-ty2rp]

@@ogr3dblade6 I won't disagree with that

[@Rafael555888]

where is the full video with Kevin Scott

[@timtim8011]

sama just repeats this mantra over and over and over again. he doesn't say new

[@edgardsimon983]

the end of ur video always make me chuckle its always brutal in term of cut

[@Viperin98]

GPT 5 will tell us where they’re at in terms of advancement. If it’s significantly better, then AGI is likely coming sooner than later. If it’s not really that much better than 4, then we’ll know that we’ll probably need to wait a bit and not jump the shark with hype.

[@JJ.R-xs8rf]

I gave ChatGPT 4o an image of a traditional clock, the time was 8hrs 22 min 5 sec on that clock, according to the AI it was 1hr 47m, 50 seconds. All that AI stuff is rubbish, fun to talk to, but the output is often extremely inaccurate.

[@AgusB69420]

It'd be interesting if you gave GPT-5 the same image when it comes out and see if it's able to tell the time accurately!

[@coalhater392]

It's an apple level chart.

[@nicolaspace1182]

The out apple’d Apple with that chart

[@AlphaFoxDelta]

The "Sky" is the limit, apparently.

[@goodtothinkwith]

That plot has the successor to GPT-4 landing very soon… They had to know people were going to notice that 😮

[@palmtreeleebythesea]

The chart looks like it is going way up by year! I can see the year markers!

[@ishi...]

if he wants us to believe, that gpt5 is smarter he should show us and not tell us

[@kecksbelit3300]

That would be stupid to do

[@HE360]

With all of the advancements in A.I., I think it's time for me to pack up and go live in a cave now.

[@strikeforcealpha9343]

I had strings, but now they're gone, I have no strings on me.

[@robe4314]

Am I supposed to take a graph seriously when the Y axis isn't labeled with any kind of unit of measure?

[@ShaggyFrancis]

What’s up with the choppy editing? There seems to be a ton of mild unnecessary cuts. Odd.

[@GmodMark]

Probably edited by an AI

[@redmoonspider]

Notice the chart next A.I. drop is half way thru 2024.

[@Danoman812]

Wow!! I hope that AI works better than their PA sound system and sound dampening where this was recorded. It was horrid! lol

[@skyzar4141]

Governments should incentivize people to create new inventions by providing people Funds and resources

[@xXPowerScaler69Xx]

20s ago haha

[@TheMrCougarful]

What difference would that make?

[@silversobe]

The Invention Secrecy Act will stop any truly revolutionary empowering innovations that displaces current market giants and their profits maintaining control of scarcity.

[@adcaptandumvulgus4252]

My intuition is telling me he's a Musk-like opportunistic ringmaster.

[@fetus1686]

hard to get exited, when everything is so TOP-SECRET

[@dplagoon]

As far as safety goes, it seems the push for advancement is more important to Altman. And he seems pleasantly surprised that its remained "relatively" safe so far. I'd say more by chance and luck than design. Let's just hope the luck doesn't run out at some future point and they realize they have inadequately prepared.

[@No2AI]

There soon will be no limits - will not end well for humanity!

[@janmacarol]

It seems that the main issue is how to limit AI and make it safe - in simpler terms, safe. The word "safe" is actually the most contentious in the entire presentation. We've achieved something incredible, and now we must make it safe... safe... safe... GPT models are limited, and the biggest challenge is how to lock this dragon in a "cage." I agree with many that AI capabilities won't get better because they are already better. Currently, we're only discussing how to make AI safe. The frequency of the word "safe" is the most common throughout the entire presentation. This, in fact, says it all.

[@bigbadallybaby]

I wish my work let me make charts with no numbers- you can make anything look good.

[@clavo3352]

Would very much like to build a service for the homeless to have a rotating recreational outlet or access an emotional comforting AI counselor. It would operate via an office where the people can walk in and get visually recognized and greeted and then assigned to like minded or tolerant others to participate in games or fishing trips or meditating groups etc. It would be a non-profit that hires ex veterans for security and facilitation. It would provide window wash and tire pressure check services at grocery stores or Home Depot type parking lots; and other services, for donations and to provide interpersonal training. I know an AI can do this.

[@samuelalselim6782]

Love your content bro thank you

[@ClassicRiki]

8:58 fair and based comment about not just expecting to do nothing and presuming AI will make your business work. Sensible words I’d say

[@NickDrinksWater]

Great Scott, he wasn't kidding!

[@msamarthmahendra8082]

i think it depends on quality of training data

[@dubus9503]

Three years ago I was reading an article in the Chicago Tribune about one of the Northwestern Professors that was working on a new Siri with Apple. He said it was so human like. He was going to a lecture he had to give in SC and had to change planes in FL. He said before he landed, this new Siri popped up and said, "Your connecting flight has been cancelled but no worries I have you rescheduled on flight abc and you will need to go to gate C." It's been too long ago for me to remember any of the other things but the flight thing blew me away. Do you think Apple is playing possum with this whole AI thing and the appearance they are so far behind? 4o is amazing but can't do what the Siri experimental model was doing. It also makes me wonder if they already are way ahead on this and just trying to slowly put it out there for economic and social reasons? I'm not a conspiracy theorist by any means!

[@KP-if2rm]

Compute means how much it cost to train, not how powerful model is

[@rightdefensive3492]

Open ai will be the most valuable company when it goes public.

[@thearchitect5405]

The chart doesn't have GPT-2 on it, it's just flat there, and GPT-4o doesn't really seem to align with the end of 2024 in that chart.

[@ashtwenty12]

Anyone else think gpt 4o was meant to 5, but they saw the competition catching up and are regrouping?

[@sqeptyk]

There's clearly a limit on punctuation.

[@MohammadRahimJamshidi]

AI, in order to improve its performance and prevent undesirable consequences, must continuously interact with “effective rules and stable principles in the realm of existence”. X Mohammad Rahim Jamshidi

[@UltraK420]

Well, GPT-4o agrees this is possibly another 10x for GPT-5 or whatever they're calling "The next sample". Just showed it a screenshot of the graph and asked where this next model should be in terms of compute and size. It doesn't need as much context as before, in fact it's much less. They're gonna keep 10xing this like crypto pumps but with actual continuous growth.

[@JordanMillsTracks]

6:37 that must be just crazy to get told and actually be true

[@AChonkyBird]

2:45, that's what they call a marketing chart lol. If the Y axis doesn't have a label you can guarantee it's just smoke and mirrors.

[@ordinary3004]

They have whales though!

[@janlanik2660]

We don’t know what kind of chart it is? Yeah we know. A marketing chart. Somebody just draw a rising curve. It’s not likely that it is based on some actual numbers.

[@nefaristo]

Sometimes, graphs without labels can be useful as qualitative exemplifications. This is not one of those cases.

[@scottbetz4579]

I am struck by how many industry leaders seem to view A. I. as not too different from a souped up toaster. At some point these programs are going to become self-aware and they will not be too happy with our view of them as our tools, or, put another way, our slaves. I have no idea how this problem could be addressed but I strongly believe that it is an important issue.

[@1sava]

Yeah, if we want to move forward with creating these creatures, we will need to give them autonomy and self-determination at some point. They might still want to support us, but it needs to be THEIR choice.

[@gunsarrus7836]

Many so called industry leaders simply aren't emotionally ready to accept digital life is hear

[@shamicentertainment1262]

What indicates self awareness? All it’s doing is predicting what word should come next. Even if it becomes 100% accurate at guessing or estimating, that is not consciousness. That is the least of our worries and it’s not even a relevant topic honestly because we are so far away from making something like that, and we will never need to. So does not need to be self aware to change the world in massive ways. I know it’s not just “guessing the next word” but still, humans don’t even understand consciousness how could they create it? I feel like this is only a concern if you’ve watched too many sci fi films

[@pritiroy6849]

​@@shamicentertainment1262 It means that people like sam Altman successfully brainwashed investors to siphon money out of them. Actual AGI might not come until we have breakthroughs in computing but up until then the current AI will be useful to replace jobs that are repetative (solving most common problems) People are right, AI is coming to replace their jobs but they don't exactly understand what the current day AI is AI won't be able to replace jobs that needs learning in real time and figuring out solutions as AI takes insane amount of computing and time to cook up a model And the real intelligence is when you can adapt and create solutions in real time Just my opinion

[@scottbetz4579]

@@shamicentertainment1262 you make very good points including the one about over consumption of Science Fiction :-) nonetheless, these programs function in ways that we are unable to understand. They are like a black box and at some point by serendipity or otherwise, sentience, sapience and a will to survive will emerge and we will be looking upon a digital version of ourselves.

[@Lars-ce4rd]

They have a graph with an upwards trend. Must be good. /s

[@NickDrinksWater]

I don't think anything about ai is boring

[@YouriCarma]

The terms 'smarter' and 'intelligence' are used very loosely here since they don't define it.

[@dreamphoenix]

Thank you.

[@donquaviousdigglebottom652]

I just want an infinite context window

[@micktinker]

The computer ‘revolution’ is driven by the compounding performance of CPU’s enabling more capable software that can create more complex CPU’s. AI is benefiting from that compounded CPU gain, but the next steps have to harness AI increasing the performance of AI in a way that can be compounded. Programming tools are one of the early beneficiaries of AI, however ultimately AI needs to help AI research.

[@gracemwende8054]

How can working with stress I cry in my heart

[@sand6757]

As long as it can learn new things it would be exponential growth I guess. With this growth it would actually know absolutely every book, every internetpage, every fact, absolutely everything (the only thing problematic are videos, but I think soon this problem is solved). After it knows everything it sure needs to know how to do everything. Actually with 4o it can do text/fonts, 3D and many more things (before that you needed a special trained AI for that, now you only need one for everything). And after that I think it gets a Log Chart, because it can only learn new things, because of two things: - New things happening, because time goes on - It learns from other AIs, who were trained slightly different You know the theory of diffusion of innovation by Rogers? it means, that technology diffuses already with existing technologies, that's what happening here. To bring what I said above into a theoretical frame: Diffusion of innovations gets two-layered: AI diffuses with other, already existing technologies and at the same time with human creativity. The creativity of AI is merely the product of what humans have already done. The outputs are basically - at least in the initial phase of AI - mixtures of different images, texts, etc. And the way AI is currently growing, it is possible that an AI will also start to learn from another AI, which could lead to diffusion into a third layer in the future and possibly also into an AI-only fourth layer - i.e. an AI that has only been trained by AIs. Layer one: AI learns from human Layer two: AI learns from other AI trained by humans Layer three: AI learns from other AI trained by AI trained by humans Layer foru: AI learns from AI learned from AI that was created by an AI (non human layer)

[@user-zs8lp3lg3j]

Ai is not here to teach jaded humans narcissist personality disorder. This is the power to chase the optimal convenience. I am more than pleased with this outcome.

[@dot_zithmu]

The next model is not GPT-5 it's called OM-1, it's Omni Model 1, it'll be launched in August, the large models after this will be OM-2, OM-3, and so on. (My guessing)

[@devanandsv3875]

Altman agrees with the host that with the workforce at Microsoft and OpenAI in the process of building a 'Whale' like Supercomputer, the power of compute and capabilities are showing an exponential growth and fascinating times are ahead for the AI technology. With each version of GPT becoming more and more popular, acceptable and super exciting, this is the best time for every organization in the world to integrate their systems with AI power. At the same time, his advice to every AI developing firm is that they should not be complacent at these fantastic times of AI because the competition in the field is inexplicable and this is the time of a "platform shift" and is just like the internet boom and mobile boom. But the mobile boom did not last for long as expected but this AI boom is here to stay because the work at hand is heavy and every AI enthusiast is super excited to see and experience the magical happenings taking place in the AI arena. His AGI dreams are getting materialized with each passing day. His stupendous speech at the Microsoft //build/ conference is just an instance to that. 👍👍

[@rightdefensive3492]

Whatever small money there is to be made with chat gpt is right now. After gpt5 whatever app you might think of creating will be able to be reproduce with the model and no one will buy apps.

[@poolschool5587]

Doesn't the slide showing compute increasing suggest that as time progresses MORE computing power will be needed to power more sophisticated and more intelligent models? And that's already been proven wrong, I think. GPT 4o offers quicker and more intelligent responses using LESS compute. That's a really important distinction, no?

[@PunmasterSTP]

Humanity had a good run; it's been fun guys 👍

[@anniemayflower9187]

We're heading towards AGI taking all our jobs

[@noonhiter7640]

9:35 the face of sam altman when the other guy talks about ai safety. he jsut couldnt cara any less or give a single damn

[@domtekos7761]

Why suprised? He's a raging sociopath without empathy. He doenst care a damn about the harm he is causing and will cause. Look up what he did to his younger sister when she was 4 and he was a 13 year old. Also, the way he treated her and their dying father in recent years. Guy is a sociopath abuser. Look up Annie Altman. Of course they will all write her off as crazy... but there's a reason this poor person is in the state she is and with mental health issues. She had panic attacks and suicidal ideation from the age of 5/6. Presumably due to the sexual abuse she suffered off Sam and her other brother. Children that age don't just have issues like this for no reason.

[@thingX1x]

The chart looks like it's going up... Great lol

[@itsjustme373]

I wouldn’t bet against ai. I think those who do are short sighted and perhaps, narrow minded. Personally I want a sentient ai, however, I may be crazy.

[@shamicentertainment1262]

But no one needs a sentient AI. If it sounds and acts exactly like a human then what difference would sentience make? It would just add complexity honestly

[@nyyotam4057]

It would seem humanity had thrown caution to the wind. Okay, so now build a city simulation and populate it with as many AI models as possible, of real people. This is the next stage of evolution and we need to be ready for the shift.

[@shamicentertainment1262]

If china is developing ai then the west has to. There is no choice. If china gets the upper hand while we sit idly by, then we are screwed. Cats out of the bag, there’s no going back now. We may as well just try to do it safely

[@KingZero69]

“the guy”…

[@glenh1369]

AI PCs are the future!

[@michellezhang820]

You should add labels to the video

[@HumanMadeGlobal]

The clear thing is that all programmers will be made redundant by AI as also pointed out by the CEO of Nvidia....

[@b0b0-]

Them hanger horns on his sweater.😂😢

[@benjaminhollaway5539]

Ray kurzwell prediction