If you have been closely following the progress of Open AIcompany run by Sam Altman with neural network can now write plain text and generate original photos with amazing speed and ease, you can skip this part.
On the other hand, if you only vaguely notice the company’s progress and the growing traction that other so-called “growth” AI companies suddenly gain and want to better understand why, you might benefit from this interview with James Currier, 5 times founder and now venture capitalist, company co-founder NFX five years ago with some friends as his serial founder.
Currier falls into the category of progress followers – so close that NFX has made many related investments in “modern technology” as he describes it, and it attracts more and more attention from the group each month. In fact, Currier doesn’t think the buzz about this new wrinkle on AI isn’t so much of an exaggeration as the realization that the broader startup world is suddenly facing a huge opportunity again. first in a long time. “Every 14 years,” says Currier, “we get one of these Cambrian explosions. We had an internet in ’94. We had one around the cell phone in 2008. Now we have another in 2022.”
Looking back, this editor wishes she asked better questions, but I’m learning here too. An excerpt from our conversation follows, edited for length and clarity. You can listen to our longer conversation here.
TC: There’s a lot of confusion about creative AI, including how precise it is or whether it’s going to become the latest buzzword.
JC: I think what’s happened with the AI world in general is that we have a feeling that we can have deterministic AI, which will help us determine the truth of something. For example, is it a piece of debris on the production line? Is that an appropriate meeting to have? It’s where you define something using AI the same way humans define something. That’s largely what AI has been in the last 10 to 15 years.
The other sets of algorithms in AI are more than diffusion algorithms, which aim to look at huge troves of content and then make something new out of them, saying, ‘Here’s 10,000 examples. Can we make the same 10,001th example? ‘
They were quite fragile, quite brittle, until about a year and a half ago. [Now] Algorithms have gotten better. But more importantly, the pool of content we’re viewing is getting bigger and bigger because we have more processing power. So what happened is, these algorithms are following Moore’s law – [with vastly improved] storage, bandwidth, compute speed – and suddenly be able to create something that looks a lot like what a human would make. That means the face value of the text it will write, and the face value of the figure it will draw, look very similar to what a human would do. And that’s all that happened in the last two years. So it’s not a new idea, but it’s at that threshold. That’s why people look at this and say, ‘Wow, that’s magic.’
So it was computer power that suddenly changed the game, not a piece of previously missing technology infrastructure?
It doesn’t change suddenly, it just changes gradually until the quality of its generation reaches a level that makes sense to us. So the answer is generally no, the algorithms are very similar. In these diffusion algorithms they have become somewhat better. But really, it’s about processing power. Then, about two years ago, [powerful language model] GPT was born, as a type of in-place computation, then GPT3 was born when [the AI company Open AI] will do [the calculation] for you in the cloud; because the data models are so much bigger, they need to do it on their own server. You just can’t afford to do it [on your own]. And at that point, things really skyrocketed.
We know because we’ve invested in Company does AI-based compositing games, including “AI Dungeon” and I think the majority of GPT-3’s computation comes through “AI Dungeon” at a time.
Does “AI Dungeon” then require a smaller team than another game maker might?
That is absolutely one of the great advantages. They don’t have to spend all that money to store all that data, and they can, with a small group of people, create dozens of gaming experiences that all take advantage of that. [In fact] the idea is that you’re going to add general AI to old games, so that your non-player characters can actually say something more interesting than they do now, although you’ll get the gaming experience fundamentally different from AI to games, compared to adding AI to existing games.
So a big change is in quality? Will this technology stabilize at some point?
No, it will always get better step by step. It’s just that the difference of the gains gets smaller over time as they’ve gotten pretty good,
But the other big change is that Open AI isn’t actually open. They made this amazing thing, but then it wasn’t opened and it was very expensive. So the groups gathered together like Stable AI and others, and they said, ‘Let’s just make an open source version of this.’ And at that point, costs have dropped 100-fold, in just the last two or three months.
This is not a fork of Open AI.
All this common technology will not only be built on the Open AI GPT-3 model; that’s just the first one. The open source community is already copying a lot of their work, and they’re probably 8 months behind, 6 months behind, in terms of quality. But it will get there. And because the open source versions cost one-third or one-fifth or one-twentieth the cost of Open AI, you’ll see a lot of price competition and you’ll see an increase in models. This model competes with Open AI. And maybe you’ll end up with five, six or eight, or maybe, maybe 100 of them.
Then, on top of those people will be built unique AI models. So you can have a real AI model that looks at poetry writing, or a real AI model that looks at how you create visual images of dogs and dog fur, or you’ll have a really specialized model to write sales emails. You will have a whole class of these specialized AI models, which will then be purpose-built. Then on top of that, you’ll have all the technology in common, which is: how do you get people to use the product? How do you get people to pay for the product? How do you get people to sign in? How do you get people to share it? How do you create network effects?
Who makes money here?
The application layer where people will go after distribution and network effects is where you will make money.
What about large companies that will be able to incorporate this technology into their networks. How difficult would it be for a company without that advantage to make money on its own?
I think what you’re looking for is something like Twitch, where YouTube could have integrated that into their model, but they haven’t. And Twitch has created a new platform and a new piece of culture and values that are valuable to investors and founders, even though it has been difficult. So you will have great founders who will use this technology to their advantage. And that will create a border in the market. And while the big boys are doing other things, they will be able to build billion-dollar companies.
The New York Times ran a Piece recently featured a handful of advertisers who say the innovative AI applications they are using in their respective fields are tools in a broader toolbox. Are people being naive here? Are they at risk of being replaced by this technology? As you mentioned, the team working on the “AI Dungeon” is smaller. That’s good for the company but potentially bad for developers who may have worked on the game.
I think with most technology, there is a kind of discomfort that people experience [for example] Robots replace a job at an auto factory. With the advent of the internet, many direct mail users felt threatened that companies would be able to sell direct and not use their paper advertising services. But [after] they accept digital marketing, or digital communication through email, they may have had huge turning points in their career, their productivity increased, speed and efficiency increased. The same goes for online credit cards. We didn’t feel comfortable ordering credit cards online until 2002. But those who accept [this wave in] 2000 to 2003 did better.
I think what is happening now. Writers, designers and architects who are thinking ahead and applying these tools to increase their productivity 2x or 3x or 5x will do very well. I think the whole world will end up in the next 10 years seeing an increase in productivity. It’s a huge opportunity for 90% of people who want to do more, do more, earn more, connect more.
Do you think it was a wrong step when AI opened [open source] what is it building, what has sprung up around it?
The ultimate leader will behave differently from the followers. I don’t know, I’m not in the company, I can’t really say. What I do know is that there is going to be a huge ecosystem of AI models and it’s not clear to me how an AI model can still stand out when they’re all aiming for the same quality and it just becomes a game. play on price. It seems to me that the winners are Google Cloud and AWS because we’re all going to make stuff like crazy.
It is possible that the open AI will move up or move down. Maybe they become like AWS themselves, or maybe they start creating specialized AIs that they sell to certain verticals. I think everyone in this space will have a chance to do well if they navigate the right way; they will have to be smart about it.
NFX has much more on its website about artificial intelligence that’s worth reading, by the way; You can find that here.