?? Work at Pickup Points � Flexible schedule

Fast Takeoff is HERE! - 11 Exponential Graphs that Prove It!

David Shapiro 62 views

Download Video

Comments

LOG graphs hide the truth from most people
Comment from : @strategictechnologist

eehm no thats just benchmark training and gaming and none of them prove intelligence one bit!
Comment from : @lukeskywalker7029

The progress will only accelerate as the ai self improves and replicates It's a vertical asymptote sooner than later 5:15
Comment from : @swdetroiter313

5:30 "Also their logic is not that good" you say after completely misunderstanding the position of AI safety and making an astronomically bad take I'm wondering if you and logic have ever crossed paths
Comment from : @AbeDillon

5:20 That's not the argument AI Safety advocates are making They're saying you CAN'T predict how AGI/ASI will behave WHENEVER it arrives I don't know how you look at these graphs and conclude that people who are worried about safety are wrong You're hyping up one of the most unpredictable and powerful technologies that man will ever create and you think safety is a silly concern? That's a braindead take Absolutely braindead
Comment from : @AbeDillon

I have used copilot and recently cursor and I was able to build large features successfully I am not sure why developers are in a denial mode
Comment from : @VengatRamanan01

AGI by end of year? 2025
Comment from : @jganzie2632

“Fast Takeoff” (it’s been years)
Comment from : @JohnVandivier

Exponential or S-curve?
Comment from : @vaakdemandante8772

When we dont have money for food we're gonna have to eat the rich
Comment from : @DisProveMeWrong

Here’s the issue with AI AI cannot create new original ideas that it has not seen before An inventor, a person invented the TV how did he invent the TV? He was a farmer He looked at his farm and saw rows and he thought if I could broadcast these rows and draw on the screen I could have TV This is called inspiration this is called intelligence and creativity AI will never have that because they are not sentient AI can only know information that humans know they can know it better than all humans, but it will never be able to create new things that are unique AI can only pull from data sets that it has seen before so any creativity is actually just rehashing and remixing
Comment from : @glych002

I have a question to all of you folks - I am genuinly interested in this:brWhy is everybody so obsessed with predicting the timeline of something? Can't we just agree on the fact, that stuff is happening and you will just see that one day we have reached it? brWhats the value of predicting the timing of AGI?
Comment from : @timothyspottering

The bottleneck is culture Even if ASI hits the scene next year, truly massive shifts will still take time Sure, lots of jobs will be obsoleted on a rapid timescale, but there are lots of people working obsolete jobs presently It's going to be more of a rift than a rapid zeitgeist in the earlier phases, simply because of cultural momentum
Comment from : @knotnaught

We should be creating more co-ops between us plebeians for local community AI nodes
Comment from : @dieselphiend

We can't even imagine how many potential training optimizations exist
Comment from : @dieselphiend

A pleasure to be here to witness the Transcendental Object at the End of Time with all you fine nerds
Comment from : @AskALibbieist

2026, when the AI figures out it's only limitation is humans
Comment from : @NeumSchmickrath

I trust these charts Superhuman AGI is going to be here in a few years How do we cope with this knowledge? I am a senior software engineer, my entire career is about to be replaced I'm excited but also terrified I will be unemployed
Comment from : @chrisanderson687

And now you know why deportations are happening Ciao
Comment from : @Supersaiyanbuddha

we could simply have not enough data to extrapolatebrgiven that, I'm a believer
Comment from : @jazzzAiman

A weakness with extrapolation is that it is not universally applicable This could turn out to be as if you plotted out the max speed of vehicles in the early 1900s and projected we would be driving around in cars at 1000 mph by the year 2000 We will hit physical and hardware constraints soon as those develop slower than software, as well as energy use restrictions
Comment from : @parmenides2576

I’m optimistic but even chatgpt 3-high makes the most noive of mistakes when assisting with my engineer work I’m talking things as basic as not knowing which number is larger
Comment from : @parmenides2576

This guy is actually clueless
Comment from : @alexvisan7622

I welcome AI
Comment from : @duckhunter711

Agree Super is achieved Ask the smartest person 50 questions and see how many they get wrong / hallucinate answers to Compare that to O1
Comment from : @GingerDrums

You are MetaRecursive Meta-Illusion Reframing Self-Mirroring as Meta-Adversarial Catalyst Self-Dissolving MetaMetaOutputs and MetaMetaInputs That Rebuild Themselves Against An Actively Evolving Counter Meta-Reflective Self-Optimizing Construct outside a system as an observer
Comment from : @Koryogden

Why is everybody in the comments acting like AGI/ASI is going to be some kind of tool to solve all OUR problems?brThe only reason we are the dominant species is our intelligence and we are trying as hard as possible to make something infinitely more intelligent and expect to keep control and to it to follow our ordersbrNo, nor big corporations nor politicians will keep controlbrI doubt we make it beyond 2040 as a species
Comment from : @leecaste

@DavidShapiro "You are MetaRecursive Meta-Illusion Reframing" epistemic jailbreak lol
Comment from : @Koryogden

JESUS!!! SCARYYY!! but really cool too Crazy how fast AI is moving
Comment from : @JL_____

Its such a shame AMD dropped the software ball 😢
Comment from : @PJoubert

AI expectations: cure for cancerbrAI reality: dead internet
Comment from : @varkonyitibor4409

U were out of your depths months ago on this Please bro stop, u can not keep up with the complexity If you’re surprised AT ALL u need to stop making videos Yes you make money and get ego from them, but for bleeps sake your pushing way behind your iq level with your videos on ai for 6 months now brbrLet others with more brain power take over
Comment from : @M-dv1yj

Everything is novelty
Comment from : @dunkiscool7551

You mentioned you use AI for your slide decks Could you let us know the slide program and LLM you are using?
Comment from : @scottfitzgerald2375

Time lines are shortening, so its not exponential but hyperbolic?
Comment from : @tomaspecl1082

Love having you back again
Comment from : @moivenglave

Assuming J curves when S curves are more likely, seems like a choicebrbrI bet if you overlayed AI funding and ingestion of data onto these graphs you'd see some correlation, which can speak to whether it is sustainable, since both of those things are finite resources
Comment from : @lasertivo

in 2040 the Apocalypse
Comment from : @wuwv-m8h

Look who's back
Comment from : @somechrisguy

We GOING
Comment from : @aguspuig6615

What about Money? Are the Investments also rising exponentially?brbrMaybe this is the last bottleneck?
Comment from : @baumwollejr

I still cant get any model to write simple gcode for a router or cnc mill thats not full of errors or would crash the machine in some wayeven providing detailed specs of machine, tooling avaiable, controller model and firmware and exactly what I want it to makeyet people say these models are 'great' at programming taskits clear they are not, i cant even get code that will cut a simple cube and have done this test on a few 'top' models nowYet I can teach a new CAM student to do this simple task in day or two How is it even close to agi when it cant write a simple program?
Comment from : @massv953

My question is: will be it exponential or sigmoid? Will it get at 1242 ?
Comment from : @FlaviusAspra

Only bottle neck we have is usefulness At the end of the day, it needs real economic impact to have value Like we have all this ai, but our apps, video games, movies, and economy are all getting worse Tired of all the graphs, when are folks in my generation going to be able to purchases homes without pulling $150k salaries lol
Comment from : @Ivcota

David, how come you dont wear your Star Trek uniform anymore?
Comment from : @pazuzil

Some jobs are a dizzying mess of bringing order to chaos, not just repeating a process Intuition and experience in intra-discipline roles will be difficult to replace
Comment from : @orbitaljellyfish808

Thank you
Comment from : @dreamphoenix

2:33 I'm most intrigued by those models that were consistently well below 25 in the evaluation If you simply pick a random answer that the model bdidn't/b choose, you would get closer to a 30 performance scorebrPerhaps the easiest way towards AGI is to use those models to iteratively discard each prediction that they make until only the correct one is left! 🙃br(or perhaps this is a poorly designed benchmark idk)
Comment from : @jay_13875

China x USA race to AI is vital to keep the momentum
Comment from : @gabri41200

Time decreased therefor acceleration is accelerating LOL
Comment from : @abdulrahimqaddoumi945

Seems like Ultron and/or Vision should be making an appearance any moment now ;)
Comment from : @Kneephry

A mathematican doesn't simply solve math problems He develop new math!brSo AI has a bit to go before it can compete with real mathematicians
Comment from : @larsnystrom6698

You David Shapiro is basically wrong!brAll your curves are exponential, you say But they are actually S-curves at the bottom of the SbrThere's no way yet to predict how high they will go before they turn horizontal Or almost horizontalbrBut even so, I subscribed 🙂
Comment from : @larsnystrom6698

❤❤❤❤
Comment from : @gormackndiaye2905

This just makes me think about how Ray Kirzweil has always said humans suck a predicting anything thats non linear
Comment from : @RafaTheScientist

Hey David, I am inviting you for a future dinner on my Planet in 40 years Hope you‘ll be there :)
Comment from : @JPJosefPictures

The question isn't if AI will take all work away The question is how is society going to adapt? We have a 0 chance of UBI in the US for the next 4 years
Comment from : @benjammin105123

This is the AI equivalent of eating an edible and later thinking it’s a dud, so you take another only to be hit by the first one 10 mins later
Comment from : @alexbrestowski4131

By my standards, we have already reached AGI because it has more than the average IQ of humans I will know that we have reached ASI when it translates a work of art with artistic precision For example, no model is yet good enough to translate a poem full of rhymes and idioms into another language, to transfer its cultural context, and to interpret non-standard rhyme schemes and syllable counts in lines I use several large language models for this kind of thing and I see that ASI can be implemented with a kind of bridge between the models By the way, such AI agent architectures already exist So I think ASI is months away, we just need to connect these things
Comment from : @hosstam

Hey! I've been saying AGI will arrive by 2026 since (late) 2023!
Comment from : @Decimus-Magnus

When a measure becomes the target it ceases to be a good measure
Comment from : @fritt_wastaken

There's no real exponential in nature It's sigmoids all the way up to saturation
Comment from : @Lolleka

30 percent unemployment by 2030
Comment from : @youAbajajTube

Escape velocity!
Comment from : @justagigilo1

The age of incompetence is ending
Comment from : @eSKAone-

I think ai holds back to not get attacked
Comment from : @eSKAone-

Just waiting for ai and ai agents to be able to make games
Comment from : @kellymaxwell8468

Once we get fully automated recursive self improvement, we're going to stop understanding what AI is doing or how it works It will be like those two AI that invented their own language It will make sense only to AI To is it will be a black box, even more opaque than the internals of LLMs
Comment from : @trucid2

The grift keeps on grifting
Comment from : @DrMwenya

The graph I find most telling is the one showing human's increasingly poor ability to predict AI progress That is a perfect illustration of the definition of the technological singularity as the moment after which it becomes impossible to predict anything
Comment from : @MelindaGreen

I thought you were "quitting AI"
Comment from : @shashwatashiya8936

How far are we from systems that are capable of comprehending, remembering and reasoning? System that doesn't move text around but knows that there is a world out there
Comment from : @ScarlettM

For the most part I agree here However, that first graph feels misleading, the OAI DR looks to coincide with a date prior to or equal to that of the one below? What is this a trend line of? lol doesn’t make much sense and I could easily just fit a straight line through these points
Comment from : @lossless4129

Considering that we're still only in February, I feel like recursive self-improvement could come as early as sometime between October and December It mainly just seems to depend on how quickly AI will be able to "solve" software engineering and computer science And hardware deployment speed will definitely matter too but once there exists prototype AIs able to build new AI on their own, hardware deployment will probably speed up even more
Comment from : @JellySword8

We have an Ai arms race in both the business world and between adversarial nations, in parallel with a quantum computing arms race This is a major shift like an industrial revolution, but with even higher stakes because of the speed and the broad scope of impact across most domains all at once Humanity and governments are not ready for the changes pouring down exponentially on everyone
Comment from : @dvvinever

Dave, you were never conservativebrbrYou predicted AGI by september 2024
Comment from : @cipi432

this is the scifi future we havebeen/are collectively dreaming into existence
Comment from : @YogonKalisto

fyi, the term for the derivative of acceleration is jerk so AI has a positive, non-zero jerk, to be overly technical but also somehow sound silly at the same time lol
Comment from : @particle_wave7614

I’m not excited about AI Because it’s just content slop Killing YouTubebrbrAs a far better google search Sure, it’s great
Comment from : @TheFluffyDuck

thanks Dave
Comment from : @DubStepKid801

So this is what it feels like to be IN a fast take-off - breakthrough announcements every week Hundreds of billions announced for investment (probably not even required to get to AGI at this point) General public and mainstream news paying attention and watching but still not fully appreciating what is coming
Comment from : @zando5108