That’s LLM bull. The model already knows hangman; it’s in the training data. It can introduce variations on the data, especially in response to your stimuli, but it doesn’t reinvent that way. If you want to see how it can go astray ask it about stuff you know very well, and watch how it’s responses devolve. Better yet, gaslight it. It’s very easy to convince LLMs that they’re wrong because they’re usually trained for yes-manning and non confrontation.

Now don’t get me wrong, LLMs are wicked neat, but they don’t come up with new ideas, but they can be pushed towards new concepts, even when they don’t grasp them. They’re really good at sounding sure of themselves, and can easily get people to “learn” new “facts” from them, even when completely wrong. Always look up their sources, (which Bard (Google’s) can natively get for you in its UI) but enjoy their new ideas for the sake of inspiration. They’re neat toys, which can be used to provide natural language interfaces to expert systems. They aren’t expert systems.

But also, and more importantly, that’s not zero-shot learning. Neat little anecdote from a conversation with them though. Which model are you using?

No. AI and, what you’re more likely to be referring to, machine learning has had applications for decades. Basic work was used back into the '60s, mostly for quick things, and 1D data analysis was useful long before images (voice and stuff like biometrics). But there are many more types of AI. Bayesian networks (still in the learned category) were huge breakthroughs and still see a lot of use today. Decision trees, Markov chains, and first order logic are the most common video games AI and usually rely on expert tuning rather than learned results.

AI is a huge field that’s been around longer than you expected, and permeates a lot of tech. Image stuff is just the hot application since it’s deep learning based buff that started around 2009 with a bunch of papers that helped get actual beneficial learning in deeper models (I always thought it started roughly with Deep Boltzmann Machines, but there’s a lot of work in that era that chipped away at the problem). The real revolution was general purpose GPU programming getting to a state where these breakthroughs weren’t just theoretical.

Before that, we already used a lot of computer vision, and other techniques, learned and unlearned, for a lot of applications. Most of them would probably bore you, but there are a lot of safety critical anomaly detectors.

This actually is a symptom from the sort of “beneficial” overfit in Deep Learning. As someone whose research is in low data, long tails, and few shot learning, there’s a few things that smaller networks did better in generalization, and one thing they particularly did better (without explicit training for it) is gauging uncertainty. This uncertainty is sometimes referred to as calibration. Calibrating deep networks can yield decent probabilities that can be used to show uncertainty.

There are other tricks for this. My favorite strategies prep the network for learning new things. Large margin training and the like are a good thing to look into. Having space in the output semantic space (the layer immediately before the output or earlier for encoder decoder style networks) allows for larger regions for distinct unknown values to be separated from the known ones, which helps inherently calibrate the network.

Which end? The main story is just a narrative device, in fact you shouldn’t really obey the narrator at all. Calling any end “The End” doesn’t make sense in the context of the game, really. Unless you just broke out of the mind control facility three times then called it quits? That end is supposed to be non enticing so that you try literally anything else before putting it down. I think the going insane end sticks with me the most. Although the game dev commentary in the recent release is fun.

Science is pushing the bounds of human knowledge. Science is only science if it propagates, otherwise it’s just someone’s discovery. Science has to be built upon, even if it’s disproven, that means it was documented well enough to be built upon. That’s not to say everything that’s disproven is science, because crackpot theories don’t often push the bounds of human knowledge.

I hope the brilliant students get their knowledge out there. (But that is unfortunately hard in academia. Despite us living in what should be a post knowledge scarcity society, we clearly aren’t.)

This is why the machine learning community will go through ArXiv for pretty much everything. We value open and honest communication and abhor knowledge being locked down. This is why he views things this way. Because he’s involved in a community that values real science.

ArXiv is free and all modern science should be open. There were reasons for publications in the past, since knowledge dissemination was hard, and they facilitated it. Now the publications just gatekeep.

This is a fair question. But also, we’re talking about one of the most influential minds in deep learning. If anything he’s selling himself short. He’s definitely not first author on most of them, but I would give all my limbs to work in his lab.

Autoincorrect.

I’ve noticed a lot of things that are considered autistic in the states specifically may be normal practice in various cultures, having worked with people in Germany, and from a large swath of Asia.

It interests me a bit, but I think the takeaway is that autism tends to manifest in a number of quirks, and the ones that don’t align with the current culture the autistic person is in are the ones that are paid attention to. That and there tends to be a bit more obsession over said quirks than in those cultures, sometimes to the detriment of the autistic person or their social life.

https://en.m.wikipedia.org/wiki/Silverfish are also pests. They eat books.

I love discord, for what it’s for. Quick synchronous talks you will never refer back to again. So not software development where indexable logs of information are necessary. I know discord has indexing, and now some form of forum. But every discord I’ve been to for development (especially modding communities) has a large corpus of synchronous logs where people get annoyed if you ask a question that was answered one before a long time ago with extremely common language making it nearly impossible to search for because the keywords have been used out of context of your question hundreds of times since the question was asked.

If the Dev communities used the forums mode in discord more, it wouldn’t always solve it, but it’d be much better. There are better places than discord for these things, but I have been trying to meet people where they’re established.

The term originated in religion, not government. And it has no definition that limits it to government creation. It’s not even limited to politics.

Not only is this not obsolete, it’s close to biographical as it closely references the first and second Artificial Intelligence Winters. The first being in the 60s. We’ve been working on these for a long time, so 5 years is short. It took until GPGPU to kick into full gear and some clever insights to get Deep Learning up and running (somewhat attributed to work published in 2011) to start reliably on this problem, and even that is an oversimplification of the timeline and the scope.

Others have mentioned oddities like the difficulty of subject matter (picture contains a bird vs picture of a bird) but there are a lot harder problems that are trivial to humans and counterintuitively incredibly hard for computers.

Designing good UX is harder than designing good UI is harder than writing good code. As a machine learning engineer, I will never be able to design UX. I have made a pretty UI once though.