Chapters (Powered by ChapterMe) - 00:00 - Coming Up 00:41 - Intro 01:09 - Googles internal view of AI 03:13 - Early at Google: Envision of AI 04:55 - Power of AI: Spelling Corrector 08:34 - Why Google isn't the dominant AI company in the world? 11:57 - OpenAI spun out of YC 13:41 - Google DeepMind Acquisition and YC Research 14:31 - Open-Source Models 16:09 - OpenAI founding story 18:10 - Why OpenAI worked? 20:56 - Control vs Freedom: Zuck the hero of Open Source 22:21 - Meta is not making money on this 23:31 - How will we get large models if it costs a billion dollars to train 24:51 - Why Zuck is incentivized to keep funneling money into open source? 29:31 - How is AI and AGI going to break in the next few years? 31:33 - Why is AGI not the unanimous view amongst smart people? 34:40 - Bookface Future: We won't distinguish a knowledge Worker 38:25 - How AI relates to Geopolitics 42:10 - Tipping point for Doomers or Optimists 44:02 - Human ego will get in the way of AI progress 44:45 - YC and Startup community has a role to play 45:46 - United Health Care Group blocking AI calls for getting claims cleared 47:36 - AI enables successful startups without massive team 48:16 - Outro
Yann LeCun is a realist when it comes to the limits of GPTs because he actually built the transformer engine that GPTs use. He’s been building models for the last 30 years, so he understands AI winters and hype cycles. People thought they could build AI in the 1970s! Every time, it seems like we're just around the corner, only to realize that we're not even close. That being said, our tools are becoming increasingly intelligent, which is still worth millions of dollars.
Great interview! I wish you had lingered more on the AGI topic. In particular I'd love to hear people speculate a bit more about how many and what key components are still missing in order to build AGI. Hearing support for building more open and accessible systems makes me hopeful for the future.
Great episode. I would have liked to hear about their view on why OpenAI is not open source and whether they think Sam will eventually flip OpenAI back to being completely open.
Not today sama. There are no idealists in market foxholes. Anyone that has outside investors certainly knows this. This is also painfully obvious to anyone competing for survival against others with non "freedom-pilled" strategies. This is literally the reason why MZ is worth billions and Craig N is worth fractions of that. Taken a step further. Cynical and foolish is the grifter that repeats the mantra "we're all about tech freedom" or "we just want to take AI where society wants us to take it". "Risks are socialized. Profit is privatized." This is the way.
Would love to hear an AMA format. While I think I agree things should be open source, there are some questions around safeguards I haven’t heard nuanced discussion on. It feels almost bipartisan. How do we enable open source AI without enabling authoritarian governments to use them for bad? How do we support open source while having safeguards around the things we don’t want to see in society? Creating bombs, viruses, misinformation campaigns, etc.
I am fascinated by the thinking “sound” Paul makes that is also so distinct to Paul Graham (PG). Never heard anyone else use it, and they are both genius people. Lucky coincidence? 🤔
One day, people will finally stop calling AI AI, and call it what it actually is today - SC. Simulated coherence. That'll temper the fantasies a lot of people, even companies, have.
@@burgerbobbelcher They do so in this context, since AI was fundamentally designed to simulate biological neural networks in an artificial environment. You can understand the meaning or play with semantics to dismiss it if it makes you feel less special.
@@cesar4729 No, they do not. Simulating biological neural networks does not in itself result in 'intelligence', it results in a convincing pretense of it. Words have meanings, you playing fast and loose with the jargon doesn't change what it actually is - not intelligent. "Playing with semantics" is the refrain of every cultist who doesn't care about being correct.
I like PB but freedom has nuance, because one person's exercise of personal freedom can impinge on the freedom of someone else, and he doesn't seem to acknowledge that.
![🔴 ฟุตบอลแชมป์กีฬา 7HD แชมเปียน คัพ 2024 [รอบ 8 ทีมสุดท้าย] - ช่วงบ่าย](http://i.ytimg.com/vi/M0S_ReCIkxg/mqdefault.jpg)
![[Part 1] รักมาไม่ทันตั้งตัว❤ หนุ่มสาวโสดถูกจับมาแต่งงานกันกะทันหัน แถมเธอยังเป็นหัวหน้าของเขา?!](http://i.ytimg.com/vi/y2pOOMkuFI0/mqdefault.jpg)
Chapters (Powered by ChapterMe) -
00:00 - Coming Up
00:41 - Intro
01:09 - Googles internal view of AI
03:13 - Early at Google: Envision of AI
04:55 - Power of AI: Spelling Corrector
08:34 - Why Google isn't the dominant AI company in the world?
11:57 - OpenAI spun out of YC
13:41 - Google DeepMind Acquisition and YC Research
14:31 - Open-Source Models
16:09 - OpenAI founding story
18:10 - Why OpenAI worked?
20:56 - Control vs Freedom: Zuck the hero of Open Source
22:21 - Meta is not making money on this
23:31 - How will we get large models if it costs a billion dollars to train
24:51 - Why Zuck is incentivized to keep funneling money into open source?
29:31 - How is AI and AGI going to break in the next few years?
31:33 - Why is AGI not the unanimous view amongst smart people?
34:40 - Bookface Future: We won't distinguish a knowledge Worker
38:25 - How AI relates to Geopolitics
42:10 - Tipping point for Doomers or Optimists
44:02 - Human ego will get in the way of AI progress
44:45 - YC and Startup community has a role to play
45:46 - United Health Care Group blocking AI calls for getting claims cleared
47:36 - AI enables successful startups without massive team
48:16 - Outro
Totally unrelated, but PB is looking shredded. Good for him!
Yann LeCun is a realist when it comes to the limits of GPTs because he actually built the transformer engine that GPTs use. He’s been building models for the last 30 years, so he understands AI winters and hype cycles. People thought they could build AI in the 1970s! Every time, it seems like we're just around the corner, only to realize that we're not even close. That being said, our tools are becoming increasingly intelligent, which is still worth millions of dollars.
This is great we need more conversations like this out in the open with others and start getting people together
Amazing content here. It's surprising to see this level of discussion in public. Tks for that YC
Great interview! I wish you had lingered more on the AGI topic. In particular I'd love to hear people speculate a bit more about how many and what key components are still missing in order to build AGI. Hearing support for building more open and accessible systems makes me hopeful for the future.
Great episode.
I would have liked to hear about their view on why OpenAI is not open source and whether they think Sam will eventually flip OpenAI back to being completely open.
Bring Eric Schmidt to the conversation next! 🛸
Love Lightcone Garry! 🏆👏
the fact that yc is freedom-pilled makes me optimistic
Not today sama. There are no idealists in market foxholes. Anyone that has outside investors certainly knows this. This is also painfully obvious to anyone competing for survival against others with non "freedom-pilled" strategies.
This is literally the reason why MZ is worth billions and Craig N is worth fractions of that.
Taken a step further. Cynical and foolish is the grifter that repeats the mantra "we're all about tech freedom" or "we just want to take AI where society wants us to take it".
"Risks are socialized. Profit is privatized." This is the way.
PB glow up is insane
definitely one of the best videos, thanks!
I is definitely the most comprehensive YC Podcast on the future of AI!
Incredibly inspiring! I think I’ll get back to building now
This was interesting & refreshing, almost felt like I'm watching one of the Lex Fridman episodes.
There is a company ahead of OpenAI’s 4o model that is cooking behind curtains. They care about country dominance over popularity and profitability.
What company could this be?
What do you mean by behind curtains? Do you mean that their company is unknown? You don't mean anthropic, right? Anthropic is obviously known.
every country is doing one though
Great interview , I wish you had lingered more on the AGI topic , speculate a bit more accessible systems makes me hopeful for the future !
"Gmail creator Paul" is such an awesome first name.
o1 released since this vid. Time for a Part 2!
Would love to hear an AMA format. While I think I agree things should be open source, there are some questions around safeguards I haven’t heard nuanced discussion on. It feels almost bipartisan.
How do we enable open source AI without enabling authoritarian governments to use them for bad?
How do we support open source while having safeguards around the things we don’t want to see in society? Creating bombs, viruses, misinformation campaigns, etc.
Is AGI the ultimate SISP?
Great sober view of AI.
Great content, as always.
I am fascinated by the thinking “sound” Paul makes that is also so distinct to Paul Graham (PG). Never heard anyone else use it, and they are both genius people. Lucky coincidence? 🤔
Hi Meena, are you reading the comments?
Never grab a popcorn so fast 😅
Lol
Awesome
One day, people will finally stop calling AI AI, and call it what it actually is today - SC. Simulated coherence. That'll temper the fantasies a lot of people, even companies, have.
Artificial and simulated mean the same thing, only one term is less uncomfortable for the fragile human ego. Which says a lot in itself.
And here I thought it was actually called Apple Intelligence
@@cesar4729 No, they do not mean the same thing.
@@burgerbobbelcher They do so in this context, since AI was fundamentally designed to simulate biological neural networks in an artificial environment. You can understand the meaning or play with semantics to dismiss it if it makes you feel less special.
@@cesar4729 No, they do not. Simulating biological neural networks does not in itself result in 'intelligence', it results in a convincing pretense of it. Words have meanings, you playing fast and loose with the jargon doesn't change what it actually is - not intelligent. "Playing with semantics" is the refrain of every cultist who doesn't care about being correct.
In the future, AI becomes smart enough to teach how to build a bomb, creator's reaction is "I'm not liable"...
Bomb making tutorials already exist on the internet
RIP Susan ❤
Cool.
why does zuck where that chain tho ? 😆
I like PB but freedom has nuance, because one person's exercise of personal freedom can impinge on the freedom of someone else, and he doesn't seem to acknowledge that.
Can I know the lady's name?
Diana Hu
47:45 Who's Malik?
@xbluebells Did you just make that up? Literally nowhere online says this monster of yours is a thing.
The winner in the long run is going to be nvidia.
I got here at 666 views
Don’t be evil lol, just joking 😅