3 ms·

Tay was released (and taken off line) in 2016. In the same year, 2016, Google announced that its Neural Translation Engine, at the heart of Google Translate se

Tay was released (and taken off line) in 2016. In the same year, 2016, Google announced that its Neural Translation Engine, at the heart of Google Translate services had developed its own "interlingua" capable of acting as a bridge between any language pair among about a hundred languages. Still in the same year, AlphaGo became the first AI system to beat a human master at Go. In 2018, BERT, the first Large Language Model amazed AI specialists with its ability to perform language tasks it had not been explicitly trained on, like question answering and machine translation. Around the same time, CEOs of large tech companies began predicting the rise of the autonomous car, that would soon be circulating in every street in the entire world (by 2019, according to some, 2020, at the latest according to others). In 2020 AlphaFold, a DeepMind AI system, repeated AlexNet's ImageNet feat, but this time in the CASP competition on protein folding. In 2021 the years of the AI image generators began, with Dall-E, StableDiffusion, Midjourney, and friends. In 2022 DeepMind announced that their AlphaCode code generator was better than 50% of some arbitrary group of programmers in an online competition. These are just some of the highlights, and excluding the GPT-x releases. Far, far from "everyone" quietly backing away and not mentioning AI, the AI hype has been increasing monotonically, and quite rapidly at that, all the way to the present day.

rsynnott 3y ago
So, by hype, I'm talking about "omg, this is going to change the world, throw all the VC money at it immediately". _That_ died away, progress continued, a new hype cycle started, at some point it will die away, progress will continue, repeat ad infinitum. Arguably, for AI in particular, this has been going on for about 50 years; see https://en.wikipedia.org/wiki/AI_winter https://en.wikipedia.org/wiki/AI_winter It does feel like they're getting closer together, in that there've been at least three AI-ish hype cycles in the last decade (self-driving cars/computer vision stuff more generally, Microsoft Tay era, current one). Read into that what you will. I do find it interesting that people are actually using the _term_ AI more with this one; for a long time there was a certain squeamishness about that.

rsynnott 3y ago

So, by hype, I'm talking about "omg, this is going to change the world, throw all the VC money at it immediately". _That_ died away, progress continued, a new hype cycle started, at some point it will die away, progress will continue, repeat ad infinitum. Arguably, for AI in particular, this has been going on for about 50 years; see https://en.wikipedia.org/wiki/AI_winter https://en.wikipedia.org/wiki/AI_winter It does feel like they're getting closer together, in that there've been at least three AI-ish hype cycles in the last decade (self-driving cars/computer vision stuff more generally, Microsoft Tay era, current one). Read into that what you will. I do find it interesting that people are actually using the _term_ AI more with this one; for a long time there was a certain squeamishness about that.