AI bubble, me thinks not, or is history a predictor?

There are too many results, but all of history.

Jon Peddie

The AI revolution, particularly in generative AI (GenAI), is marked by its ability to uncover patterns and create novel content, such as art, literature, and music. While GenAI offers a higher-resolution view of data, it is based on historical information. The value extends across the supply chain, but it’s akin to driving by looking in the rearview mirror. GenAI, though powerful, is not genuinely creative or inspired; it reconstitutes existing data. The AI boom’s sustainability is uncertain, and while it may face limitations, its transformative impact on various industries is undeniable.

Seeing all the fervor about AI, the announcements of new, specialized AI chips, the flow of reports of discoveries and predictions of AI takeover, and the boost in P/E ratios and share price of any company with AI in its PR, one could conclude that the AI revolution, as it’s being called by some, is indeed the most significant thing to happen since the telephone or the discovery of X-rays.

The miracle of AI, and generative AI specifically, is how it can be used to find things (patterns, anomalies, obscure differences) and create new and unusual things such as art, literature, music, videos, and gaming scenarios.

It seems like magic and is seen as miraculous. GenAI has given us a new, higher-resolution view of our data. Things that were not observable before are being found. And that is most certainly highly valuable. The value extends through the entire supply chain, from the raw data to the software, processors, and reports. What we used to call data mining has taken on a new life and meaning. It’s like the people and organizations that find gold in the tailings of a spent mine or an abandoned rivulet of a creek.

But, it is like driving by and looking in the rearview mirror. Everything we are learning from GenAI is based on history. The age of that history is a function of the investigator’s investment in its Web crawler or data miner. It takes time and resources to review all the data available. The free version of OpenAI doesn’t look beyond 2021, so anything later than 2021 is not considered. For more current examinations, you must pay. That’s not an unreasonable request. That OpenAI will allow us to use it as we do for free is a fantastic gift to humanity. Somebody paid for the hundreds of processors used in our free use of OpenAI, Dall-E, and other LLM AI engines.

We should remember the parable: Give a thousand monkeys a thousand typewriters, and one will write Shakespeare. Is GenAI nothing more than that? One could argue no because it doesn’t just reproduce something like a sonnet from Shakespeare; it creates something new and original. If you give OpenAI or Dall-E the same request, you will get a different result every time. That’s the G part of GenAI—the feedback on itself.

However, it’s all from history. It’s not genuinely new or novel. It’s not inspired, and it is not creative. It is simply reconstituted. And yes, the weighting can be tuned, and amazing things can be found in our data that we didn’t see before and might never have seen. Stars and nano-sized tumors that were invisible to us can be found. Yet, they were always there. Just because we couldn’t see them doesn’t mean they suddenly materialized. It’s a little like Schrödinger’s cat—is it alive or not?

The magic of what AI can find and even create will continue. We have so much data to crawl through, and we have done a pretty good job of saving a lot of it, even with the Fahrenheit 451 episodes in our past and currently. Also, as GenAI becomes more accessible, new private data sources will be exposed. And, whereas the creatives have been initially shocked and frightened by the things GenAI has produced, already we are seeing the more imaginative of that sector exploit AI as a tool and not complain about it being a threat.

So, is the AI boom a bubble? Will it fade out as the great new panacea after its limitations have been realized? Probably, such is the cycle of things. But it will take a while—years. We are at the tip of the beginning of the leading edge of what we can do with AI. And like any new technology, there will be miraculous discoveries and developments benefiting humanity, and unimaginable, horrible destruction will be made possible by it. And laws cannot prevent that—it’s childish to think they can; if laws could stop things, there would be no wars or crime. But it will keep the politicians’ names in the papers and on TV, and that’s all they are about.

The boom in share prices of companies in the AI realm will experience the bubble pop first. That will take less time; there is not an infinite demand for AI tools and processors anymore than there is an infinite amount of money to buy such things. So, the demand will flatten. Once organizations get an AI system, it will be reused. The AI system market will follow any other market and peak, and then ease into a replacement market—PCs and smartphones are current examples. TVs and cars are another. One of the signals will be when AI start-ups stop getting bought up. When funding for AI start-ups flattens or vanishes, then the party is over. As of now, there is no sign of that. Check back in a year from now.

In the meantime, allow your awe to blossom as you read about the new AI processors being developed by Microsoft, Google, Baidu, and a half dozen start-ups. Allow your wonder to bloom as you read about researchers’ discoveries using GenAI. The AI boom will go through a cycle like all others before and after it, but this one will truly bring amazing benefits to us all.

When will AI predict the end of AI?