We call it AI because no one would take us seriously if we called it matrix multiplication seeded with a bunch of initial values we pulled out of our asses and run on as much shitty data as we can get our grubby little paws on.
Beating neural approaches to sentence classification for out-of-domain datasets with ... nearest neighbour over gzip (finding the ratio between the compressed size of the documents combined vs. smallest compressed size of the inputs being compared).
This is both hilariously simple, and somewhat damning of how the other methods do in this context.
I have run out of severance and now I'm running out of savings. It's do or die now, folks. Job offers, job leads, job hunting advice...please send them all my way, and/or boost for reach. Freelancing is on the table, too. Details about me and what I'm looking for, to follow.
Anybody out there looking for an ML or software engineer with >30 years total experience and ~20 years in the industry?
I have extensive experience with #Python and #ML frameworks, particularly #TensorFlow, and I've worked on #NLP and #ImageProcessing both in the workplace and in personal open source projects. My resume is available here:
sounds like the much heralded job of the future, "prompt engineer" is no longer needed 😅
"Battle and his collaborators found that in almost every case, this automatically [AI generated] generated prompt did better than the best prompt found through trial-and-error. And, the process was much faster, a couple of hours rather than several days of searching."
There's this Google internal document, for example, that points out FLOSS community is close to eating Google's and OpenAI's cake:
ttps://www.semianalysis.com/p/google-we-have-no-moat-and-neither
So here is my question to you:
What are the best examples of useful, small, on-device models already out there?
A study (not yet peer-reviewed) done by a team of researchers from the Swiss Federal Institute of Technology concludes that between 33% and 46% of workers on platforms like Amazon's Mechanical Turk had used AI models to automate work typically done by humans needed to train AI systems.
Justice systems around the world are using artificial intelligence (AI) to assess people with criminal convictions. These AI technologies rely on machine learning algorithms and their key purpose is to predict the risk of reoffending. They influence decisions made by the courts and prisons and by parole and probation officers.
We made a free, open course on #Neuroscience for people with a #MachineLearning / quantitative background. Get deep in #ComputationalNeuroscience or just indulge your neurocuriousity. I believe that neuro and ML can learn from each other and do better together than on their own.
The course has 34 short videos from introductory topics right up to recent discoveries we still don't fully understand. We also have practical exercises focussed on open ended discovery, fully compatible with Google Colab.
My thanks to brilliant co-developer @marcusghosh, and contributors @GabrielBena, Swathi Anil and Greta Horvathova.
Over the next year, I'll be turning this into an 'interactive textbook' with videos, text and runnable code in one place, and welcoming contributions on new topics, corrections, etc. through GitHub issues. All our materials are freely licensed for reuse in your own courses too.
Why this new course? There's a lot of intro neuroscience courses out there, and a lot of ML for neuroscientists, but I wanted this one to be specifically for quantitative people who are curious about the brain, how it might be similar and different to ML.
We really need a different term than ‘AI’ to describe private, on-device machine learning that benefits individuals but (a) doesn’t violate anyone’s privacy, (b) doesn’t destroy the environment and, (c) doesn’t enrich smug Silicon Valley tech douchebros like Sam Altman.
Please boost for reach if this kind of stuff interests you. Will post more on this later.
Once upon a time, there was a cool emulator frontend called Retroarch. This emulator wasn't accessible until I and a few other gamers went to them and asked about adding accessibility. An amazing person known as BarryR made it happen. Now, if you turn on accessibility mode in settings, or pass the "--accessibility" (or something like that) flag on the command line, you get spoken menus, including the emulator's pause menu, good for saving states and such. Then, using PIL and other image processing Python utilities, running a server and hooking into Retroarch, the script allowed players to move around the map, battle, talk to NPC's, ETC. The only problem was, no one wanted to test it. The blind gaming community pretty much spoke, saying that we want new games. We want cool new, easy accessibility. So that's what we have no, follow the beacon or get sighted help in the case of diablo and such. It's sad, but meh. It's what we wanted I guess. No Zelda for us. So, this is about as far as he got:
To expand on what devinprater was saying: I am working on an accessibility pack/service for Final Fantasy 1 for the NES (this was what was shown in the latest RetroArch update). The idea is similar to how Pokemon Crystal access works, but it's using the RetroArch AI Service interface to do so.
Right now, the FF1 access service is mostly done, but I need more testers to try it out and give me feedback on how it's working. Right now, you can get up to the point where you get the ship, but there's no code to deal with how the ship moves, so that still needs to be done. Likewise with the airship later on.
The service works the latest version of RetroArch, on linux and mac, but not windows. This is due to how nvda reads out the text and until the next major update to nvda (which will have a feature to fix this), it'll have to wait. If you have those, I (or maybe devinprater) can help you set it up on mac/linux to test out. The package itself is available at: https://ztranslate.net/download/ff1_pac … zip?owner=
I'm happy to announce the start of a new free and open online course on neuroscience for people with a machine learning or similar background, co-developed by @marcusghosh. YouTube videos and Jupyter-based exercises will be released weekly. There is a Discord for discussions.
For more details about the structure of the course, and to watch the first video "Why neuroscience?" go straight to the course website:
Currently available are videos for "week 0" and exercises for "week 1", but more coming soon.
Why did I create this course? Well, I think both neuroscience and ML can be enriched by knowing about each other and my feeling is that a general purpose intro to neuro or comp-neuro isn't the right way to inspire people in ML to be interested in neuro.
I hear a lot about neuroscience inspiring AI, but I think there's understandable scepticism about that from ML people. I don't want people to take neuro ideas and apply directly to ML, I just think we get a richer picture of what both fields are doing if we think more widely.
In other words, we should be thinking that we are somehow studying the same problem in different ways. You see that in the early history of the field, and it's very inspiring. (Yes, this is pretty much just saying that cognitive science is cool, but my scope is a bit narrower.)
The focus then is not on how neuroscientists think the brain works, but on the mechanisms the brain uses. These are strange, inspiring, and often their contribution to intelligent behaviour is still deeply mysterious.
The first video of the main part, on the structure of neurons, finishes with recent research (from @ilennaj and @kordinglab among others) on what the function of dendritic structure might be. No answers, just ideas.
And that's going to be another key part of this course. Research level problems are not hard to find in neuroscience, and the aim of this course is to empower students with the tools to start finding and working on them straight away.
Most of the exercises in the course won't have correct answers. They're starting points for further investigation. We'll be downloading and exploring open neuroscience datasets using methods from computational neuroscience and ML.
The course is not supposed to be comprehensive. It's a short course and the aim is more to get inspired and start on a longer road. I'd expect everyone to get something different out of it, and I'm happy if for some people their take home is "neuroscience is not for me"!
In some ways, it's the course I would have liked to get me into neuroscience and for my incoming PhD students from non-neuro backgrounds to be able to take. It's personal, and full of the sort of stuff that inspires me to be interested in neuroscience.
Well, I hope that some of you might be interested to follow along in the next few weeks, and since it's the first time I'm giving this course please do give feedback by email, Discord or however you like. Also, please feel free to re-use materials however you like.
Whenever I see OpenAI's Sam Altman with his pseudo-innocent glance, he always reminds me of Carter Burke from Aliens (1986), who deceived the entire spaceship crew in favor of his corporation, with the aim of getting rich by weaponizing a newly discovered intelligent lifeform.
I'm on the latest episode of Brain Inspired talking about #Neuroscience, #SpikingNeuralNetworks, #MachineLearning and #Metascience! Thanks Paul Middlebrooks (not on Mastodon I think) for the invite and the extremely fun conversation. For the explanation of why this picture you'll have to listen to the episode. 😉
Also, if you're not yet listening to Brain Inspired you should be - and support Paul on Patreon. He provides this free for the community with no adverts. What a hero!
Here we go again: The boom in artificial intelligence (AI) and quantum computing will drive a spike in energy use, the National Grid has predicted. Data centre power use 'to surge six-fold in 10 years' https://www.bbc.com/news/technology-68664182 all this power usage for shity LLM/AI to write poems, pictures, videos etc from stolen data to train AI. #artificialintelligence#machinelearning
A 'black box' AI system has been influencing criminal justice decisions for over two decades—it's time to open it up (phys.org)
Justice systems around the world are using artificial intelligence (AI) to assess people with criminal convictions. These AI technologies rely on machine learning algorithms and their key purpose is to predict the risk of reoffending. They influence decisions made by the courts and prisons and by parole and probation officers.