|
I tell people, if you know how this stuff works, you'll understand its a fugazi. It produces a statistically likely sequence of tokens (words) that resemble data it was built with. There's nothing intelligent about it, it cannot create anything novel. It is truly nothing more than a language model. Everything that comes out of one of these is a hallucination, in the sense we mean when talking about AI. I think it's a poorly understood form of compression, a very lossy form of compression. I think it's big corporate adherents know all that, I think, particularly after the internet and the industry it unlocked, smartphone and the internet access that they made ubiquitous, software developers are just looking for things to sell to people. All the innovations have happened already in the software world, but this big industry was built while this happened and they can't just pack up and go home. The truth is, you can't just sell software to users, you have to sell utility, and most people don't have a real need for software beyond what they already have available, so shiny new trinkets it is. They've got video streaming, communication, information retrieval, report generation, and that's it. You can automate stuff, but that's not the end user buying software, that's the end user buying a widget that software is a component of, and a part they really don't care about. A symptom of what I'm talking about can be seen in the ever increasing needless complexity people keep finding in their lives, the software industry as it applies to direct to consumer sales is very little more than hype and noise and marketing. It's empty calories. AI is just another thing in a long line of these things, yesterday it was augmented reality, day before that it was IoT, the list goes back quite a number of years. The last real big innovations that end consumers bought that improved their lives was smartphones and video streaming. Everything an end user can benefit from software wise they already have. Easy communication, multimedia content and symmetry of distribution, near free information availability, I can't think of anything beyond those 3 broad categories, money maybe if you're into the bitcoin thing. Maybe "AI" as compression can help with local information retrieval when it's more mature, which is very powerful, but the branding and marketing around it as if it is a mind is not conducive to success on that front. Software is well on it's way to being a mostly business to business thing again, like it once was, which necessarily means a big bubble is going to pop. That's where the innovations are happening, but for us regular people it's all marginal improvement of what we already have as a side with a healthy serving of hype and disappointment. |