(Roughly) Daily

“The key to artificial intelligence has always been the representation”*…

AI is coming for search. OpenAI’s chatbot offers paraphrases, whereas Google offers quotes. Which, asks the estimable Ted Chiang, do we prefer?

… Think of ChatGPT as a blurry jpeg of all the text on the Web. It retains much of the information on the Web, in the same way that a jpeg retains much of the information of a higher-resolution image, but, if you’re looking for an exact sequence of bits, you won’t find it; all you will ever get is an approximation. But, because the approximation is presented in the form of grammatical text, which ChatGPT excels at creating, it’s usually acceptable. You’re still looking at a blurry jpeg, but the blurriness occurs in a way that doesn’t make the picture as a whole look less sharp.

There is very little information available about OpenAI’s forthcoming successor to ChatGPT, GPT-4. But I’m going to make a prediction: when assembling the vast amount of text used to train GPT-4, the people at OpenAI will have made every effort to exclude material generated by ChatGPT or any other large-language model. If this turns out to be the case, it will serve as unintentional confirmation that the analogy between large-language models and lossy compression is useful. Repeatedly resaving a jpeg creates more compression artifacts, because more information is lost every time. It’s the digital equivalent of repeatedly making photocopies of photocopies in the old days. The image quality only gets worse…

Should we bank on AI in search? “ChatGPT Is a Blurry JPEG of the Web,” in @NewYorker.

For more of Chiang’s thoughts on AI, listen to (or read) his interview with Ezra Klein, in which he suggest that “most fears about A.I. are best understood as fears about capitalism.”

Also apposite: “AI, Minus the Hype” and “Imagining The QAnon Of The AI Era.”

Jeff Hawkins (who seems to be agreeing with Baudrillard that “the sad thing about artificial intelligence is that it lacks artifice and therefore intelligence”)


As we fiddle with our filters, we might spare a thought for a man whose work has created a gargantuan training set for AI: Alphonse Bertillon; he died on this date in 1914. A police officer and biometrics researcher, he applied the anthropological technique of anthropometry to law enforcement, creating an identification system based on physical measurements. Anthropometry was the first scientific system used by police to identify criminals; before that time, criminals could only be identified by name or photograph. While the method was eventually eclipsed by fingerprinting, then DNA analysis, it is still in use.

Bertillon is also the inventor of the mug shot. Photographing of criminals had begun in the 1840s only a few years after the invention of photography, but in 1888 that Bertillon standardized the process.

Bertillon’s work has been hugely impactful– and lies at the root of many AI systems being developed to finger criminals (especially via facial recognition). It’s worth remembering that his (flawed) evidence was used to wrongly convict Alfred Dreyfus in the infamous Dreyfus affair.

Bertillon’s mug shot self portrait (source)
%d bloggers like this: