ChatG BS

It seems rate rises haven’t killed all overvalued tech…by David Kimberley

 
ChatGPT has been plastered across the media since its launch at the end of November. For those not familiar with the programme, it is something of a cross between a chatbot and search engine. You ask the system a question and it will give you answer, often one that is lengthy and detailed.

The bot has already been banned by educational authorities in New York, after pupils were caught using it to do their homework. Software nerds have commented on how it can help them with coding. And it has unsurprisingly spawned a swathe of articles on whether or not we’re going to be replaced by artificial intelligence.

Less commented on has been the fact that ChatGPT regularly says stuff that is total nonsense. For instance, software engineer Phil Mulholland has written about how ChatGPT made up code, which he described as “a mishmash of examples of an obsolete feature taken incorrectly out of context, glued together with a plausible seeming but imaginary and impossible feature.”

Similarly, Christopher Graves, the founder of the Ogilvy Center for Behavioral Science, used ChatGPT to look for scientific studies. The chatbot (or whatever you want to call it) made up a bunch of fictional papers, including titles and author names, instead of just admitting it didn’t know any – perhaps making it more human than if it had actually known them.

Use the bot for even a small length of time and you’ll quickly get answers like this. For example, if you say it takes one car two hours to drive from London to Birmingham, ChatGPT will infer that it takes two cars four hours to do the same trip.

The “plausible seeming but imaginary” snippet of Mulholland’s article is thus an apt description of much of the chatbot’s output. In short, a large volume of the text ChatGPT spits out is a kind of verbalism – it is good at giving the impression, to the uninitiated reader, that it knows what it is talking about. Dig even a tiny bit deeper and you often realise that it’s either false or that nothing of any substance is actually being said.

This is not true across the board, nor should it take away from just how impressive some of what the bot comes out with is. But the response to it has brought back memories of the market bubble we saw over the past couple of years.

Reporting from news site The Information claims that Open AI is lossmaking and had revenue in the “low tens of millions of dollars” in 2022. And yet recent reports suggest the company is looking to raise capital from investors at a $29bn valuation. If we are generous and say the company made $60m last year, this would mean it is ‘worth’ almost 500x sales (not profits!). Reports from Proactive Investor also indicate there is going to be a tender offer to the tune of $300m, meaning existing shareholders of the company will be exiting if it goes through – not exactly a resounding endorsement of that valuation.

The lesson of the last few years should have been that Utopian promises from lossmaking businesses, backed by ultra-low interest rates, should be taken with, at an absolute minimum, some level of skepticism. Have we already forgotten carmaker Nikola pushing its truck down a hill to show it ‘driving’ to investors? Perhaps investors may want to look at executives at ‘crypto’ exchange Coinbase, who have collectively sold more than $1bn (that’s fiat currency) in stock since its IPO. And who could forget Oatly? The oat milk producer said in its IPO prospectus that it would be tapping into the “accelerating growth and influx of new consumers to the plant-based dairy market.” Valued at $10bn when it went public, or about 5x global sales for the entire oat milk sector in 2020, the company is still deeply unprofitable and has lost almost 90% of its value since its first day of trading on the Nasdaq.

This is not to say that ChatGPT is destined for the same fate. Who knows? Maybe the company will improve its product and it will become a tool we use all the time. But the current round of ‘investment’ should serve readers as another reminder of what the excesses of the past few years led to.

Or as ChatGPT might put it:

“A high price-to-revenue ratio can indicate that a company’s stock is overvalued. In general, a ratio of 500x revenues would be considered extremely high and may indicate that the company’s growth prospects are being heavily overpriced by the market.”
 

 
Past performance is not a reliable indicator of future results
 

Disclaimer

This is not substantive investment research or a research recommendation, as it does not constitute substantive research or analysis. This material should be considered as general market commentary.