Large Discussion: <a href="https://news.ycombinator.com/item?id=41263143">https://news.ycombinator.com/item?id=41263143</a>
> Brynjolfsson: ”I asked Sundar [about Google losing the initiative]. <i>He didn't really give me a very sharp answer.</i><p>> Schmidt: Google decided that work life balance and going home early and working from home was more important than winning. The reason startups work is because the people work like hell. I'm sorry to be so blunt, but the fact of the matter is if you go found a company and compete against the other startups — like [we did] in the early days of Google — you're not going to let people work from home and only come in one day a week.”<p>It’s pretty cack-handed to publicly talk about your successor like that. While the WFH part of this got a lot of press I wonder if the free-wheeling side-swipe at Sundar Pichai had more to do with the cringe backpedaling.<p>Also, the humblebrag about his medal!… he’s an investor!… he showed Sam Altman his calculations that OpenAI will need lots of electricity!… he’s an investor!… he wrote a report setting national AI policy that was “only about 752 pages long”!… he’s an investor!<p>Schmidt has done some amazing things and his achievements will eclipse many others but I do wonder if even he feels a bit of the post-FAANG blues where one misses the glory days of ones peak, over performing and telling everyone about it to show you’ve still got it.
Full transcript is available on GitHub:
<a href="https://github.com/ociubotaru/transcripts/blob/main/Stanford_ECON295%E2%A7%B8CS323_I_2024_I_The_Age_of_AI%2C_Eric_Schmidt.txt">https://github.com/ociubotaru/transcripts/blob/main/Stanford...</a>
Really fascinating to see such a mix of blind spots, insecurity and, yet, good insights on full display here. It’s almost as if he’s a normal human being at the end of the day, despite his circumstances.
This thing just gets repeatedly uploaded, I guess deletion is not that useful now.<p>Put that aside, Eric Schmidt did told the truth (maybe except AI and NVidia stock thing) and offered the public a glace of how things actually work within the industry.<p>One important thing he mentioned was (after simplified) "just steal, and deal with justice later", which actually mirrors what Steve Jobs once said "Great artist steal". This should remind everyone, that if you want to have significant achievement in this industry, you at least needs to guard against thefts, because there WILL BE (maybe ALREDY are) people who comeout with a "be shameless or be dead" mindset after been educated by people like that man.<p>(Also, I found it quite funny that one of Eric's girl friend has probably took his lesson too)
Few notes:<p>- Alphabet have made a switch from a (mostly) tech-driven company to a pure finance-driven one, when this happen you can expect a bit of growth, even a sharp one for a little time, than an inexorable decline;<p>- the more and more centralized web means search engine became dysfunctional beasts try to pour water to specific mill instead of being open search and for a company living on search more than anything else it's a problem, of course there is GMail a project born when Google was tech-driven, but again GMail is search, so Drive, and in that "specific domain" they have substantial competitors and they can't have innovation being financially-driven now;<p>- LLM-push a way to reach a kind of dummy semantic holy grail of research while very popular is a substantial failure. Yes it have a certain wow-effect, but results are such low quality that the wow effect will not last longer and current "better-than-Alphabet" competitors in that field still have to see any meaningful and durable growth and profit.<p>Long story short the anti-remote-working is just another droplet in a substantially lost PR/élite battle to force people in the city, the sole way they have to remain alive because finance and services can only live at large city scale where they own anything and they rent/sell services to anyone. In a spread world they simply die killed by SME innovation and personal ownership that value substantial innovation against services.<p>IMVHO the big-tech model, a cleptocracy born out of Xerox tech once they have found a way to make it anti-user, is at it's end. I do not know what could happen next because so far ALL élites want such model to rule slaves, and we have lost much of intelligence, competence of the Xerox time, and no one else in the world seem to be there, China included, who was able to surge as an industrial power but still lag behind in software, even in the current sorry state of IT. But it's clear that the service model of big tech is dead. Or they found a way to reinvent or they are done.
It will be interesting to see if Schmidt’s comments about the future of AI being mostly a China-US game will come true, or if less powerful models will ultimately be more useful than single centralized powerful ones.<p>The analogy here might be between nuclear weapons and drones; the former are controlled entirely by a small number of countries, but the latter probably have more of a direct role to play in the future of warfare - and yet aren’t costly or difficult to make at all. The assumption tends to be that the most powerful well-capitalized tech always wins, but I don’t think that is necessarily the case.
> The rich get richer and the poor do the best they can.<p>> The fact of the matter is this is a rich country's game, right?<p>You heard it straight from the wolves mouth folks, don't act surprise when they bite.
it's funny how both threads here on hn, the top comments focus on his quote badmouthing work life balance.<p>when anyone with half a brain can see that work life balance has much less impact on a tech company than replacing the tech founders with the likes of him and a sea of business major middle Managers...
Eric ran Google while the Android team often worked 7 days a week in office (famously: “bacon Sundays”) and still managed to blow every advantage they had over iOS by making repeatedly terrible strategic decisions.<p>But yeah, he’d have won AI by keeping butts in seats…
Listened to the first few mins. Did he say anything controversial in the? The part about Google work-life balance isn't that big a deal. It's no secret that startups grind more.
"The country is going to have to learn critical thinking. That may be an impossible challenge for the US." (ts 30:57)<p>"Impossible" is a bold qualifier. Maybe he was exaggerating - maybe not. Will critical thinking necessarily become more difficult - to learn, teach, exercise - with AI? It's possible it will help people become better thinkers, but I don't think that's a guarantee.
<a href="https://en.m.wikipedia.org/wiki/Eric_Schmidt" rel="nofollow">https://en.m.wikipedia.org/wiki/Eric_Schmidt</a><p>See the "Public positions" section of the above article.
What is Google supposed to win at? Being the most popular search engine? Being the highest grossing ad agency? Having the highest average salaries? Selling the most used cars? Curing cancer? Owning more generators than anyone else? Contributing money to youth groups? How do we judge whether Google is winning?
At about 39 minutes in, he's asked if efforts analogous to seti at home can be used to get around the scaling problems with training the next round of big models.<p>He gives a strongly NVidia oriented answer that I happen to think is dead wrong. Pushing more and more GPU/Memory bandwidth into more and more expensive packages that are obsolete after a year or two isn't the approach that I think will win in the end.<p>I think systems which eliminate the memory/compute distinction completely, like FPGA but more optimized for throughput, instead of latency, are the way to go.<p>Imagine if you had a network of machines, that could each handle one layer of an LLM with no memory transfers, your bottleneck would be just getting the data between layers. GPT 4, for example, is likely a 8 separate columns of 120 layers of of 1024^2 parameter matrix multiplies. Assuming infinitely fast compute, you still have to transfer at least 2KB of parameters between layers for every token. Assuming PCI Express 7, at about 200 Gigabytes/second, that's about 100,000,000 tokens/second across all of the computing fabric.<p>Flowing 13 trillion tokens through that would take 36 hours/epoch.<p>Doing all of that in one place is impressive. But if you can farm it out, and have a bunch of CPUs and network connections, you're transferring 4k each way for each token from each workstation. It wouldn't be unreasonable to aggregate all of those flows across the internet without the need for anything super fancy. Even if it took a month/epoch, it could keep going for a very long time.
I have a question out of curiosity: Has anyone or anything ever succeeded in deleting/removing anything from the Internet once it goes up?<p>I’ve realized that anything that goes online will be (perhaps unintentionally) leaked, hacked, uncovered, archived, altered, restored, referenced, and what not.