The $1.5 Trillion AI Leap of Faith: Are We Investing in the Wrong Future? | Utilizing AI Episode 47

So this week, we are continuing a conversation we recently had and sort of taking it to the logical conclusion, the leap of faith, as it were, because all of us, in fact, are taking AI to the logical conclusion with a leap of faith here.

Essentially, all of society, now you may object to me saying this, but I'm going to say it anyway, all of society has decided to invest huge amounts of money and huge amounts of time and energy and resources and attention into building AI, into building something that has been promised to be artificial general intelligence, or as some people are saying, super intelligence.

And the question is, are we actually building building that? Is that actually what we as society have decided to build? And are we actually getting any closer to that? So, Guy, I think I'm going to throw it to you first.

Tell us a little bit about the leap of faith that we as society have taken.

Well, I think it's very understandable.

Let's start with that.

When I first encountered the current wave of AI, that would be the large language models, late 2022, the famous ChatGPT 3.5 release by OpenAI, it was clear and obvious that this was something that could be helpful, helpful to me, helpful to my company.

You type in, you know, just something in plain language and you get out something that at least looks really helpful.

And I think that the models as they have matured and, you know, the harnesses and the services and everything as they have matured and the agents that have been built, that has shown itself again and again and again.

There's two parts to this, though, that make it leap of faith that has resulted in what a recent report from Bain counted as $1.5 trillion globally being invested in AI.

The first one is something that I just harp on all the time, which is that the way generative AI, which is what most people encounter as AI, the way it is trained is is sort of unconsciously to fool you into it being true and good and useful and helpful.

The training data set is what you might call, you know, various masses of human produced stuff, but it's not like the models understand them.

The models are simulating the same kind of output as they consumed during training as input.

And that necessarily logically means that there's no one or thing verifying that it's true necessarily other than ultimately the human user and reviewer of it.

So that's one thing.

It's designed to look compelling, helpful, truthful.

So the leap of faith in that sense is unconscious.

You're being made, you know, you're being compelled into this leap of faith with AI.

And the second element of it is just all of the hype that has surrounded this because people are making lots and lots of money.

And I mean retail folks who have invested.

are making lots and lots of money.

I mean, managers who are getting bonused for whatever metrics they're producing, they're making lots and lots of money.

So I'm intentionally leaving out the whatever, hundreds of people, billionaires globally making even more billions.

I think that calling it a leap of faith is the right thing because we just don't realize the religion in a sense, that we have joined here and the faith that we are putting into this market.

And that Bain report, which I'd love to have a chance to summarize, you know, the Bain report, I think, represented a good faith effort by a really smart and resourceful organization, namely Bain, to try to actually figure out all the numbers here.