Generative AI: Hype or Reality?

It has been 6 ½ months since ChatGPT launched, and some people are starting to question whether the excitement around Artificial Intelligence - and Generative AI in particular - is really warranted.

Could it be that we’re in a huge crypto or dot com-style bubble?

Evidence in favour of the bubble hypothesis

A few recent signs have emerged that seem to imply we are, in fact, towards the top of a FOMO frenzy-fuelled mountain of hype:

Exhibit A: GenAI Start-up Valuations

The big news this week was AI Startup Mistral securing funding at a valuation of $260M.

So what? Building large language models needs a lot of cash!

The catch? Mistral is only 4 weeks old, there’s no product or customers, and apparently it’s first workers only started work a few days ago (non-paywalled).

It does have some talented founders, but racy valuations like this without any proof of concept can be a prime sign of bubble dynamics in a market where investors are so overcome by fomo that they throw cash at anyone who mentions “dot com” “blockchain” or “AI”, regardless of whether there’s a proven business case.

Exhibit B: Share prices

It’s not just the startups…

Chipmaker Nvidia’s share price has gone to the moon:

With a PE Ratio of over 200X (more than 5X S&P 500 average), investors who bought Nvidia stock today could see Nvidia increase profitability 10X instantly and would still be waiting 20 years to get an ROI.

Exhibit C: ChatGPT Popularity

What about ChatGPT? After all, it was ChatGPT’s release back in november 2022 that got this AI hype started.

Well, after initially setting records and becoming the fastest-growing product ever, some people are questioning how many people actually still using it.

A recent Morgan Stanley Research survey found that “only 19% of respondents have used it.”

Google Trends data seems to imply that interest is tailing off, after peaking around the time of GPT-4 launch:

So… AI Hype bubble then?

Well, hold on…

Sure, “only 19%” of Morgan Stanley Research’s survey respondents had used ChatGPT. But that’s still more than Apple’s annual smartphone sales marketshare, or about as many people in the US who used the internet in 1997.

Actually, 19% user penetration for a brand new product in just over 6 months is fairly extraordinary - particularly given there has been no advertising and the product was still in Beta.

What the realtime data tells us

But there is also plenty of data that seems to imply Gen AI is having a real world impact right now.

AI is already present in US jobs data, attributed to 3,900 losses in May. This is 5% of all job losses, and more than those caused by downturn or loss of contracts.

We are already hearing about companies replacing their workers with GenAI tools like ChatGPT.

But even those that aren’t replacing their workers with Gen AI are seeing a benefit from the technology, with studies demonstrating productivity increases of between 14% (in the case of customer service agents in this study from the National Bureau of Economic Research) and 55% (in the case of coders using Github Copilot).

And any technology that is making employees more productive and/or saving companies money will continue to grow because investment will pour in.

And we’re seeing that:

Even if we cautiously assume that a lot of this highly speculative investment cash is ultimately lost, we are still left with hugely talented and well-funded teams moving this technology forward.

The real inflection point may not be here yet

When GenAI is built into the tools that the world uses every day through tools like Google Duet AI, or through Google search (SGE)... we may reach an important inflection point in usage.

At the moment, use of Generative AI requires people to go to a different website, learn new software or (in the case of Midjourney) go on some horrendously un-user friendly learning curve.

But with companies like Adobe, Google and Microsoft all building GenAI tools into their market-leading products, people will no longer need to alter their behaviours to use GenAI.

That makes this unlike any prior technological innovation, where user behaviours had to change or consumers had to buy a physical thing like a wireless radio, router, smartphone.

So our analysis is that the data shows short-term hype in valuations, absolutely. But it also shows that AI has a long way to go, and is already having a real-world measurable impact.

Any tech that saves companies money, makes their employees more productive AND improves customer satisfaction has a long road ahead of it.