#44: The Next AI Winter

Ian Bowie
Hello and welcome to AI unfiltered with me Ian Bowie, and our resident expert, Michael Stormbom, where we will be talking about everything to do with AI in our modern digital society and what the future holds for all of us.

Michael Stormbom
So the topic would be is there going to be a new AI winter.

Ian Bowie
I actually hadn’t realized that there’s been a first winter.

Michael Stormbom
So let’s briefly recap what the first winter was about. So actually, there’s been several time periods to which the term AI winter has been applied with two major episodes in particular that lasted for years. So one in the 70s and one in the late 80s and early 90s. So the AI winters were essentially the periods that followed after AI research had hit a wall and failed to live up to the hype with unrealistic expectations and not yielding enough return on investment and resulting then in less interest in AI research and funding drying up. So in the UK, for example, large scale research was pretty much stopped during the 1970s. So I think that’s the… basically, plenty of stuff happens in AI research, but eventually just hits a wall.

Ian Bowie
No, I can believe that. That’s going to happen. Yeah. Oh, oh, they asked the question. Is this really necessary and why are we doing it?

Michael Stormbom
Yeah, indeed. And that leads me to another article. Which was on LinkedIn. And, because there are these annual reports on the adoption of AI, and actually, from last year, the adoption of AI had actually dropped, so less companies were looking into looking into AI and machine learning than the new previous year, which would tend to suggest that maybe companies are realizing that this is not the end all be all. You have to make huge investments and then maybe it still doesn’t do what you want it to do. Either because you don’t have the right data or the right expectations or understanding of your use case. And I’ve previously been on the record saying that I believe that there will be an AI bubble coming. Right because there’s been so much hype put into AI. There’s so much hype there’s so much money that has been invested in various companies claiming to do fantastic things, and then it just doesn’t, it doesn’t really yield any income and eventually it will all come crashing down.

Ian Bowie
Right. Well, I mean, it all makes sense to me. Yeah, I don’t see why there wouldn’t be another so called AI winter. I think there’s a limit to what people need from AI. Governments on the other hand, that’s a different matter. You know, we’ve talked a lot about facial recognition and security and, and of course weaponizing it for…

Michael Stormbom
But I think also there can be a backlash that when people realize what these things are actually being used for.

Ian Bowie
Yeah but I don’t think governments care.

Michael Stormbom
Possibly not. Maybe we are a little bit less apathetic about it.

Ian Bowie
They always sort of throw out the card of, well, you know, we’re keeping you safe.

Michael Stormbom
In the name of security, you can justify just about anything, yes.

Ian Bowie
Going back to the AI winter idea. And there are two sort of areas here. There’s what can be done for people in general. And there’s what governments want to be done.

Michael Stormbom
Yeah. I think I’m thinking more of the AI winter more in the sense of the dotcom bubble of the early 2000s. Because I mean, there’s so much, I mean, at some point in time, you could just you just put it you have AI in your… and then you get your venture capital money. So in that sense it’s a little bit similar to the dotcom bubble, there’s so much hope and hype and…

Ian Bowie
Yeah, but then on the other hand, of course, can they afford this to become a bubble that will burst?

Michael Stormbom
You mean? Is it too big to fail?

Ian Bowie
Well, well, you got to wonder haven’t you.

Michael Stormbom
You do, because I mean so much… maybe it is too big to fail. Yeah. At the end of the day.

Ian Bowie
Yeah. You know, the dotcom thing, I don’t think

Michael Stormbom
That was in the early days, in a way. Yes.

Ian Bowie
Yeah. I think this this has just become enormous.

Michael Stormbom
Now it’s, yeah, I mean, of course, it’s a completely different society that we live in, it’s so ingrained in everything we do. So you might be right, it just, it can’t be allowed to fail.

Ian Bowie
They’ll just throw more money at it, if anything. Yeah.

Michael Stormbom
And I mean, especially if you’ve seen the, like the media uproar around chatGPT which we discussed in the previous episode, chatGPT and what it produces, which in the best of circumstances is quite okay. But it can also produce complete…

Ian Bowie
Yeah, no, but the question there was, you know, are we producing something that we really need? Yes. I think we’ve got to a stage where it’s so big, they can’t afford it to fail. So that the answer will be well, we don’t know. But let’s throw more money at it to try and find out. Yeah. And it’ll just go on and on and on.

Michael Stormbom
Until there’s no more money.

Ian Bowie
There will alwasy be money because they’ll just print more.

Michael Stormbom
True.

Ian Bowie
They’ll never be a shortage of money.

Michael Stormbom
Can I have some?

Ian Bowie
I should declare myself a sovereign nation and start printing my own money.

Michael Stormbom
You know, there was the party, the party here in Turku that wanted Turku to be an independent country.

Ian Bowie
Is that right? Doesn’t surprise me.

Michael Stormbom
Like okay, sure.

Ian Bowie
you can do that apparently.

Michael Stormbom
You can declare all sorts of things. I can declare myself Pope if I want to. Doesn’t mean anyone will pay attention to it

Ian Bowie
Pope of what? Yeah. Well, you’ve got the right name anyway, Michael, I mean, it’s very a saintly name, isn’t it? Yes. Yeah, absolutely. I’d have to use my second name, David. But you actually live opposite your own church.

Michael Stormbom
I do do… well, you can’t say where I live here in undisclosed location.

Ian Bowie
Close to.

Michael Stormbom
Yes. At some distance from from a church.

Ian Bowie
Yes. My own church. Yes. Named after me. Yes.

Michael Stormbom
I might as well go the full distance and declare myself Pope. Yes. No, but the point being is you canf declare all sorts of things if you want, but it doesn’t mean that anyone will pay attention.

Ian Bowie
No, unless you declare them on GPT3.

Michael Stormbom
Yes, in which case it’s absolutely true.

Ian Bowie
Write an article about why I am the Pope. The true pope.

Michael Stormbom
Yes. Yeah. The anti-Pope. Yes.

Ian Bowie
Yeah. The Pope that you should be sending all your money to, as we speak. Yes. Going back to an AI winter. I don’t think so. I think it’s going to be an AI summer for quite a long time to come.

Michael Stormbom
Yeah. You might well be right here to set up when I see. I see numbers like, okay, that the number of companies that are adopting AI are dropping and I think there’s maybe not as consequential a bubble as what happened with the dotcom one. I see some a little bit yeah

Ian Bowie
but do you think that companies that are involved in developing AI seriously developing you know, like, Deep Mind and open AI and those kinds of companies, so there’s so much money invested in companies of that nature? They can’t afford to let it fail. I just don’t see it, because I could see a domino effect of collapse if that happened. Because I think actually this AI concept is so much bigger than the dotcom bubble ever was.

Michael Stormbom
That is true, and it’s more ingrained in our society. Yes, we are more digitized than ever. I do think though that because a lot of companies are you know, “we should do something with AI”, is the approach but they have no real thought into what

Ian Bowie
But they’re just idiots.

Michael Stormbom
Yes, but there’s plenty of idiots.

Ian Bowie
There are. I mean, going back to that thing, somebody said to me the other day, you know that if you could die from being stupid, that be quite a lot of people wouldn’t be here anymore.

Michael Stormbom
So I mean, if your company you don’t have your data in order, you can’t necessarily do anything sensible with AI either.

Ian Bowie
But then, of course, what’s going to happen is those companies who are struggling, somebody clever will come along and say, Well, I will help you with this. You pay me lots of money, and I will sort it for you.

Michael Stormbom
True, but the question is, will that sorting it out. Will it actually yield so much value? Additional value? This is the question.

Ian Bowie
You don’t know do you, I mean that the consultants live on that uncertainty, don’t they?

Michael Stormbom
Sure. But I mean, just the general idea of adopting AI will it actually yield so much or so many benefits that it’s been hyped to be it and I think that’s I mean, he’s more leaning towards that people are overestimating what you can do with it.

Ian Bowie
Well they are, but I mean, I think, for example, going back to my comment about consultants, I think people have overestimated what you can do with them for donkey’s years, and they still make hundreds of millions.

Michael Stormbom
Yeah. And if you’d like to hire us for a project…

Ian Bowie
So no, I’m not a big fan of consulting because….

Michael Stormbom
You could use GPT to be the consultant.

Ian Bowie
You could easily do that. Yeah.

Michael Stormbom
Well, I mean, we’ve come to, we’ve spoken about this concept of virtual workers. So basically, some system that does an automated thing, but then it also has a persona and you can talk to it. So you can of course, plug in GPT there and then

Ian Bowie
Well, I was gonna say I actually think that GPT three perhaps, is going to be one of those things that you can take it you can feed in a load of data and you can ask it for a solution, for example.

Michael Stormbom
I think that would be the.. The basic idea of GPT3, not the, like the general use case, but the specific use cases.

Ian Bowie
Specific. Yeah, yeah, I think that’s the way it’s gonna go. Yeah, that you kind of buy the basic thing.

Michael Stormbom
And put in your own data.

Ian Bowie
Yes, yes.

Michael Stormbom
Yes.

Ian Bowie
Yeah. Yeah. It’s a bit like an engine, you know, you buy a certain kind of engine block, and then you build your actual engine around it.

Michael Stormbom
Or you buy a tool, or a piece of software or whatever.

Ian Bowie
Yeah, exactly. Yeah. Yeah. So it’s like a template.

Michael Stormbom
In a way. Yes.

Ian Bowie
Yes. Yeah.

Michael Stormbom
A foundation. Yeah.

Ian Bowie
Yeah. Yeah.

Michael Stormbom
And other metaphors. Yes.

Ian Bowie
So then, you know, we’ve actually solved the problem haven’t we for GPT3.

Michael Stormbom
Yes. So that one’s sorted then. I mean, I think that’s generally true of these large language models. I believe it’s more more useful for these like specific use cases, rather than what is maybe now in the mainstream consciousness, which is this idea of AI general solving, but now you can take it and you plug in your own data, and then you can solve specific problems with it.

Ian Bowie
It’s like image generation, you know, you’ve got a company and okay, we need to have some marketing images or, or a specific image for a specific purpose. And rather than hiring an artist or a graphic designer or an advertising agency or whatever, they just, you know, type in the sort of key instructions to an image generator, and it creates it for them.

Michael Stormbom
And actually, we can use the same approach. There’s a for example, the stable diffusion models that I’ve been playing around with so people have taken those models and added their own images and trained with that. So fairly specialized models so for example, you can generate images in the style of Tron the movie, okay. Interesting.

Ian Bowie
Yeah. So there you go.

Michael Stormbom
So you can you can take the same approach if you like, we have a specific sort of images we want to generate so and we have this data that we can feed into so we create our own model.

Ian Bowie
So actually, this idea of climate change and global warming hasn’t really hadn’t potential effect on the so called AI winter.

Michael Stormbom
But it’s an interesting thought, will it contribute to the AI winter or the possible AI winter eventually if if the carbon footprint of these AI solutions become unacceptably large?

Ian Bowie
Well, of course I mean, it sounds a bit like an oxymoron to me because if you think about how much energy AI uses, it must be a net contributor towards climate change and global warming.

Michael Stormbom
I think so as it stands. Yes.

Ian Bowie
Yeah. So the idea that something that creates so much heat could then turn into a winter or am I just being silly here.

Michael Stormbom
No but indeed. I was just looking up his article about the fact that it’s getting increasingly difficult to reproduce AI results. Since Dr. Jeffery Funk here on LinkedIn, I’m reading an article from him here. And that was then, especially in the realm of healthcare. No the example here was about using algorithms to detect lung cancer and in the research papers, so they had attained 90% accuracy. But when you then try them in the real world, their accuracy was actually significantly lower. So then it was in the 60% to 70% range. So this is the reproducibility problem that you have these articles you have these research showing us fantastic results, but then when you try to replicate them, they are in fact not as good, nowhere near as good. But that’s just one, one example. And I mean, that comes back to the topic of the AI winter, that there’s so much hype, and then when you actually put these things to the real world, then it doesn’t quite live up to the hype. And causing disappointment, and…

Ian Bowie
You know, the thing is, I mean, all right to talk about an AI winter I mean AI is so broad and involved in so many different areas of life and industries, etc, etc. That, okay, perhaps in in in the medical healthcare industry, it might fail or it might slow down in terms of its development. So I think there’s not going to be I don’t think there’s going to be an AI winter per se. In the field of AI. There might be an AI winter in areas in which AI is being utilized.

Michael Stormbom
Yeah, and I think in healthcare in particular, it’s been very hyped up the use of… hasn’t really lived up to

Ian Bowie
Hasn’t delivered yet.

Michael Stormbom
Yeah. And I mean, it should be noted that of course, there have been AI winters in the past and they have been followed by an AI summer. Yeah, well, so I mean, just because…

Ian Bowie
As it should be I mean, surely, you know, summer follows winter

Michael Stormbom
Indeed. Yeah, even in the AI world.

Ian Bowie
And winter follows summer, or autumn. Yeah, yeah. Perhaps we’re entering an AI autumn.

Michael Stormbom
Yes. Yeah. Yeah, the leaves are turning, turning red. What was I about to say? Yeah, so I think that might be I mean, we have spoken about these large language models and chatGPT. I think at the moment there’s so much hype attached to it. Yeah, that that type will surely die down.

Ian Bowie
Yeah, well, yeah. Well as it does in most things, I mean,

Michael Stormbom
It’s how it dies down, that’s the key determiner whether…

Ian Bowie
So it’s a bit like electric cars. You know, I mean, some of them look really cool. But are they actually useful? As a, you know, everyday means of transport outside of city areas

Michael Stormbom
And as a way of mitigating climate change.

Ian Bowie
And definitely as a way of mitigating climate change.

Michael Stormbom
But I mean, I think it’s a question of how the hype dies, right? So if there’s… that will determine whether it’s an AI winter or not. And of course, as we discussed, there’s so much money invested in…

Ian Bowie
Well, I was about to bring that up. Yes. The hype will only die if the money runs out. And as you have said yourself, there’s already so much money involved in this. You know, money is a massive driver, isn’t it? Okay, there was a good example. There was an article I read this morning, and it was about a giga factory, you know, one of these battery factories that they were supposedly going to launch in the UK in Northumberland, and the company behind it, they’ve just gone into receivership. Because the money ran out.

Ian Bowie
Because, obviously, there’s not enough interest from investors. Yeah. So that’s it. So there’s a battery winter for you. And yeah, this is quite interesting because everyone’s hyping up electric cars and electric this and that and all the other and yet, they’re not willing to put money into something which is fundamental to electric vehicles, batteries, and battery production. So I just wonder, you know, also, if there’s a little bit of sort of rethinking that is this really the future and if the same thing happens in the AI world, then yeah, the money will dry up and there will be a winter and it’ll be very cold.

Michael Stormbom
A cold day in hell. Yeah, yeah. Yeah.

Ian Bowie
I don’t know. Yeah.

Michael Stormbom
Yeah, I mean, we’re just speculating. So we don’t know what the future…

Ian Bowie
We’re speculating. And of course, you know, like I say that the AI world is so broad and it touches so many different things that there might be a winter in one specific area. I mean, I think, all right. What about these VR goggles

Michael Stormbom
VR goggles.

Ian Bowie
I mean, that seems to have slowed down quite a lot.

Michael Stormbom
Yeah, that was a couple of years ago. So yeah. I was quite interested in getting a pair myself, but, you know, I’ve tried them out and I mean, they’re just not at the level where you would want to wear them on your head. They’re heavy, like you can feel the pressure on your bridge everyone else and the image quality isn’t yet so good. You get a headache after a while so I mean yeah, it’s not quite at the level where I’d be willing to spend money on it.

Ian Bowie
And will it ever be. And do we want it to be I mean, who wants to walk around with goggles on their head? But no, VR doesn’t do it to me. And that Hells Bells? I mean, can you imagine sitting in an airport lounge, waiting for your flight, and you’ve got all these people wearing these ruddy VR goggles, and doing all kinds of weird things.

Michael Stormbom
Sounds like a recipe for..

Ian Bowie
Yeah, or on the aeroplane, you know? Not? Yeah. No, thank you.

Michael Stormbom
I don’t think there.., I think it’s a nice thing at the moment. Yeah. Yeah. We didn’t ever be anything more than that. Not unless they get significantly lighter on the head. Yeah, no, definitely.

Ian Bowie
Yeah, no, I just…

Michael Stormbom
So in today’s episode, we are dissing VR glasses.

Ian Bowie
That’s right. So no, I think the idea of an AI winter is that there will be perhaps a bit like we have now, areas where it gets cold and it snows. But there will still be areas where the sun is shining, and it’s nice and warm.

Michael Stormbom
Yes. And after winter comes summer.

Ian Bowie
Yeah, well, that’s true. Yeah.

Michael Stormbom
Yeah. So we’ll see what happens.

Ian Bowie
And it doesn’t look like the money’s gonna run out does it

Ian Bowie
Not for the moment at least.

Ian Bowie
You’ve been listening to me Ian Bowie, and my colleague Michael Stormbom. On AI Unfiltered, and for more episodes, please go to aiunfiltered.com Thank you.

Transcribed by https://otter.ai