Hacker News Reader

Comment by Xcelerate | Hacker News Reader

I find the current VC/billionaire strategy a bit odd and suboptimal. If we consider the current search for AGI as something like a multi-armed bandit seeking to identify “valuable researchers”, the industry is way over-indexing on the exploitation side of the exploitation/exploration trade-off.

If I had billions to throw around, instead of siphoning large amounts of it to a relatively small number of people, I would instead attempt to incubate new ideas across a very large base of generally smart people across interdisciplinary backgrounds. Give anyone who shows genuine interest some amount of compute resources to test their ideas in exchange for X% of the payoff should their approach lead to some step function improvement in capability. The current “AI talent war” is very different than sports, because unlike a star tennis player, it’s not clear at all whose novel approach to machine learning is ultimately going to pay off the most.

View on HN

In fact this is much more optimal when looking at history. Strangely, success often comes from dark horses. But it makes sense, since you can't have paradigm shifts by maintaining the paradigm. Which is what happens when you hyper focus on a few individuals (who you generally pick by credentials).

The optimal strategy is to lean on the status quo but also cast your net far and wide. There's a balance of exploration/exploitation, but exploitation feels much safer. Weirdly you need to be risky and go against the grain if you want you play it safe.

With the money these companies are throwing around we should be able to have a renaissance of AI innovations. But we seem to just want to railroad things. Might as well throw the money down the drain.

> If I had billions to throw around, instead of siphoning large amounts of it to a relatively small number of people, I would instead attempt to incubate new ideas across a very large base of generally smart people across interdisciplinary backgrounds.

I had an interesting conversation with an investor around the power vs knowledge dynamic in the VC world and after a few hours we'd basically reinvented higher education with reverse tuition. Defining a general interest or loose problem space and then throwing money over a wall to individuals excited about exploring the area seems wasteful until you look at the scale of failed projects.

Agreed, and I suspect the explanation is that these plays are done not to search for a true AGI, but to drive up hype (and 'the line').

The higher the line goes, the higher the expected value of return on investment. It’s a virtuous cycle based on a bet on all horses, but since the EV is so high for first mover advantage for AGI, it might be worth it to overleverage compared to the past for your top picks? These are still small sums for Zuckerberg to pay personally, let alone for Meta to pay. This is already priced in.

There are quite a few assumptions in your message. But here is, I think, the most crucial one:

> but since the EV is so high for first mover advantage for AGI

is it? why? I cant see why this should be the case. where exactly do you think the "moat" for AGI will come from?

At this level, you probably don't need a moat to recoup your surplus sunk costs due to AI talent acquisition. You just need a good day in the market, likely the same day you announce you've achieved AGI. It's kayfabe accounting.

Any announcement of AGI will be immediately controversial. Valuation increases will depend on whether people actually believe it and what they're able to sell. Expect public opinion to be fickle and share prices to be volatile.

Decent odds we see some pretenders make that announcement before the real deal. A company with the real deal would probably make bank, but I don't pretend to know when that will come or who that might be.

I don't think that price movement is necessary to make money as an outsize shareholder, especially during high volatility. Zuck knows how to buy the market leader early, so we might have already seen the creator of AGI be poached by Meta, they just haven't realized what they know yet until they work with others in the org.

It's possible that they won't be the early bird to catch the AGI worm, but sometimes the investment squeeze required to be the first mover isn't rewarded in market juice, especially if the second mouse can use your AGI to create their own AGI cheese.

We saw it happen already with Deekseek.

SV has already thrown it down the memory hole but for a good three months, until everyone else copied their paper, the SOTA reasoning model available to the public was open source, Communist[0] and came out of a nearly defunct Chinese hedge fund.

[0] If you don't believe the communist part just ask it about the American economy.

yep. and even the sports analogy doesn't fully explain what's going on. if we are talking "true" AGI with potential to replace people wholesale their strategy is telling in that they aren't optimizing for the "end game". maybe it's a factor of just gathering all the mindshare/hype/resources and THEN they can go actually figure it out /s.

it would be like if you were looking to train the next tennis star that had the ability to basically upend the entire game as we know it. maybe you saw a few people with a unique way of playing that were dominating an order of magnitude higher. you DEF would see teams and coaches having open tryouts and trying very unconventional things for anyone they could find that had promise.

for the record i think "AI" is not hype and is changing the way things are done permanently, but it's yet to be seem whether all these spent billions can actually meet the expected return (AGI). it's hard to separate out the true innovations from the obvious grift/money grab also going on.

Reply on news.ycombinator.com