Uh oh. The inmates are running the asylum again.
Let’s face it, most technical people just aren’t great marketers. The best scientists, researchers, and engineers just think differently than the best communicators. In fact, most aren’t so great at communicating with other humans at all. If you take offense at this statement, you probably aren’t in that cohort.
But we do have a small problem.
AGI. Gen AI. LLMs. ML. RLHF. WTF.
Society at large is anxious about Artificial Intelligence. Probably because smart technical people have created a technology that can communicate better than good communicators themselves. But one of the big issues is also how we talk about these technologies, because as we can see from the machines, how you talk about something is very important to how effective you can be at communicating. The machines are now pretty good communicators, and that has us worried.
Society has taken to calling everything ‘AI’. ‘What’s our AI strategy?’, ‘What are we doing about AI in education’?, ‘Will AI replace our jobs’. This overly broad moniker muddies what we’re talking about.
So I think we should get rid of Artificial Intelligence. Not the technologies, but those two words.
No one calls a calculator AI*, yet why not? We generally think of people who can calculate numbers in their head as very intelligent, and few people can match the performance of a calculator. So isn’t a calculator artificially intelligent?
Well, the term Artificial Intelligence inherently raises the question ‘What is Intelligence’. That is a very interesting, yet pragmatically useless question. Kinda like ‘What is the Universe’ — a question that both makes me stare in awe at the stars and break out in an existential sweat. So how should we refer to this new stuff?
A large language model is a type of language calculator, with some differences:
a) It’s not based on logic, but on prediction
b) It’s not based on a set of programmed rules, but what has been written previously
c) This language calculator has an additional interesting property in that a type of logic gets encoded into the prediction model itself. Meaning that in reading the predicted word output, humans can interpret a synthetic logic in the output. Put in a loop or layered approach, that means this language calculator can perform logical steps.
These things evolve, and I’m sure we’ll converge on a name. Charles Babbage called his 1837 invention an Analytical Engine. That’s catchy. A hundred plus years later and society settled on Computer. That is, until we put it in our pocket. Then in the US we called it a mobile phone. Despite the fact that no one actually uses it as a telephone*.
The issue is that the folks building this latest set of technology really want them to be general purpose. This is something of an obsession with computer science types. Just ask anyone who codes if they’re writing a framework for something — then be prepared to listen for two hours. Yet these models aren’t general quite yet, which is confusing a lot of non-computer science types. Right now LLMs are something of a Word Predictor served with a side of OMG-it-looks-like-logic-but-we’re-not-sure-exactly-how-it-works.
So I propose we bring in the marketers, or just use the LLMs to name themselves, and start referring to the current batch of technology in a more understandable frame. Some suggestions:
a) Autonomous Back Propagated Variable Weight Language Predictor
b) Multi-layer RGB Pixel Generator
c) Patch-based Iterative Solving Multi-Frame Concatenator
Oh, I forgot to mention that I studied computer science and am currently working on a generalizable framework. But my friend (who BTW spent a large portion of their education learning about 17th century dutch painters and insists strings are for tying things) has suggested shortening those to Language Predictor, Image Creator, and Video Generator. Not nearly as precise, but shorter is more efficient. Also an obsession of computer science types.
Whatever we do, let’s stop using the acroynm AI. Otherwise we’re not really being fair to the people who designed your washing machine. After all, it is quite intelligent.
*Unless you are raising money for a calculator startup
*Other countries are more creative than the Americans. The Germans have been calling a mobile phone a ‘Handy’ for years. That name is so broad it can mean anything. And this is coming from a culture that doesn’t skimp on precision. Unsere deutchen Freunde probably used a marketing department
