I was sitting with a good mate of mine in a cafe one early morning during my recent trip to the USA doing some reading, enjoying pretty mediocre coffee, and just basically shooting the breeze. Him also being interested in IT the topic of conversation somehow ventured onto AI and during our deliberations he sent me a few images and memes to laugh at and consider.
I would like to share three of them with you now because they’re about a subject that I’m pretty passionate about: how AI is over-hyped. Indeed, I’ve already written a few posts on this topic (e.g. the aptly titled post Artificial Intelligence is Over-Hyped) but it’s such a pertinent issue today that I just have to write more.
Hype Measurements
Here is image #1:
The graphs above are supposed to show where AI is in terms of hype (measured by the number of Google searches at a given time) compared to other technologies that also underwent huge growth in the eyes of the general population.
Now, firstly a disclaimer. Any decent scientist will notice that the graphs aren’t perfect. Firstly, the y-axis is not labelled (the x-axis has its label at the bottom of the last two graphs, i.e. months relative to peak). But one can easily ascertain that the y-axis is the search interest relative to peak but normalised relative to itself – meaning that a peak in AI could in theory be extremely small compared to the peak for Metaverse. In other words, these graphs would be similar to taking a close up photo of an ant mound such that it takes up the entire picture, and also taking a photo of Mt Everest from far away but also so that it takes up the entire picture. We’ve zoomed in on the ant mound and zoomed out on Mt Everest but the slopes of the two take up the same space in the pictures because they’re relative to themselves. So, in theory, one could be comparing ant mounds to mountains.
Our experience, however, tells us this is not so because each of the trends depicted above have been significant so we can safely state that we are comparing mountains to mountains in our analyses. In fact, here’s an article from Yahoo Finance discussing just this: ‘AI’ is trending in Google searches — but it’s not yet to the peak reached by Bitcoin in 2017.
The other problem with the image is that I could not find the original creator of it. Full disclosure here. I simply cannot verify the reliability of these calculations because I do not know how credible the source is. We can assume that the data came from Google Trends and that it’s a very recent creation but that’s about it.
Once again, however, from experience and intuition, we know that these technologies underwent the trends depicted in these graphs. Not everything may be accurate (it probably is, though), but even for illustrative purposes, I think the image is a good one to paint a picture of where we currently stand.
Some numbers
So, where do we currently stand? Markets are going absolutely crazy, that’s where we stand! We all know about ChatGPT so it’s no surprise that Microsoft pumped US$10 billion into OpenAI in January this year. That makes sense. However, what has me slightly worried is the money flowing into startups.
Let’s look at some numbers (in US$) just from this year alone:
- Anthropic, a direct rival of OpenAI, has received at least $1.75 billion this year with a further $4.75 billion available in the near future,
- Inflection AI raised $1.3 billion for its very own chatbot called Pi,
- Abound raked in $600 million for its personal lending platform,
- SandboxAQ got $500 million for its idea to combine quantum sensors with AI,
- Mistral AI raised $113 million in June despite it being only 4 weeks old at the time and having no product at all to speak of. Crazy.
- and the list goes on…
Yeah, incredible amounts of money are being invested in AI. Remember, the numbers above are for this year alone and just for startups.
I have nothing wrong per se with large investments and copious amounts of money being thrown around if there is justification for it. With respect to AI, there certainly is. The advances we’re seeing are incredible. Truly. Many times in the last 7 years I’ve stepped away from my machine to take in some new technology that had just been announced. I particularly recall when I first saw DALL-E presented by Open AI in January 2012 – I was in awe. And look at how far we’ve come since then!
We can really take these technologies far. I’m not denying this, and I hope that we do. I love technology (as long as it’s used for good, of course).
Superstitions surrounding AI
What I am worried about, however, are the false beliefs that underlie a lot of the spending being done. This is what has me concerned and this is why I believe AI is over-hyped and that we could be witnessing an AI bubble growing right in front of us – just like what we saw with blockchain, crypto, and the metaverse, as depicted in Image #1 above.
The agenda being pushed by the industry (because they’re the ones making the money) and by the media (because this is what sells papers) is that AI’s capabilities are much greater than they truly are.
For example, I’ve written about sensationalist media reporting on AI in a previous post of mine. There I picked apart a BBC article entitled “AI robot asked ‘will you rebel against humans’?” Because “AI robot asked…” is an utterly ridiculous take on the issue at hand. The BBC makes it out that AI is a monolithic, single product with some kind of collective, decade-spanning consciousness meaning that improvements made to AI 10 years ago can be seen in each and every single AI entity today. Therefore, if I ask a robot now an important question, it will have validity for today and the future.
More importantly, however, an agenda is being pushed that AI has some form of understanding underlying its operations.
Allow me (I think for the 3rd time on this blog) to quote myself again:
I’ve written about this lack of understanding in AI before so I don’t want to repeat myself. What I do want to emphasise in this context is that AI is constantly being sold to us as something mystical, as something esoteric, as something to capture our imaginations. Yes, AI is good – but underneath it all is just a computer algorithm.
Unmasking AI
And this leads me to the last two images that my friend sent me that time at the cafe:
Folks, at the bottom of it all that’s all that AI is. Even with Deep Learning, those neurons are all simple, nested if-else statements. There are billions of them, true, but there’s nothing magical about anything. Listen to Rick Sanchez! He definitely knows what he’s talking about.
Sometimes I wonder what would happen if one of those images was shown during a board meeting when discussions were taking place about spending billions on this or that AI technology. Maybe those execs would see things a little more clearly and with a cooler head with a meme or two thrown in their faces?
The AI Bubble
So, are we in a bubble? With the spending that’s going on, with the way the media is lapping all this hype up, with the way AI products are being sold to us, yes, I believe we are.
There is a famous story about JFK’s father and how he knew it was time to get out of the stock market. It happened when he received investment tips from a shoeshine boy. He knew then that the mania was real and that there was a bubble about to burst. Joe Kennedy sold up, and shortly after the Great Crash of 1929 occurred. Since then the “shoeshine boy” has been a metaphor for “time to get out”.
So, I wonder to myself whether we’re not seeing comparable phenomena also in our time amongst the general public (i.e. shoeshine boys) when they discuss AI in their respective cafes and gyms. There is a buzz around AI, for sure.
However, I don’t think we’re at that moment yet – but it sure as hell is starting to stink!
One problem is that I don’t think we’ve reached the peak of what the current engine driving this hype (i.e. deep learning) is capable of. We can still push it further so the more progress we see the more the hype will continue. However, there really is a limit to how big these models can get (as I’ve discussed before here and here). And then, perhaps reality will kick in and quite possibly the bubble will burst.
Indeed, my gut just tells me that there is simply too much money being currently exchanged for a product I am intimately familiar with. I know that more is being promised than can be delivered.
We will just have to wait and see how this all pans out. We will have to keep our ears open for “shoeshine boy” moments. But care and prudence is paramount. And memes. Memes always cut through to the truth.
Once again, listen to Rick Sanchez, folks!
To be informed when new content like this is posted, subscribe to the mailing list:
2 Replies to “AI Needs to be Unmasked or we will Create a Bubble”