Obviously this is bad for people's #privacy and #security and everything but, I'm really excited about the #genetic#AI research that this leak could power! #ChatGPT was built on #LibGenesis, what will they build with this?
I asked #BingChat (creative #GPT4) to write a nursery rhyme about a billionaire-owned social network struggling to cover a major news event in the Middle East.
Note that "struggling" was the only word that might have guided the #LargeLanguageModel in such a dark direction. Or perhaps "billionaire" also carries a not-insignificant quantity of negative connotations for the #GenerativeAI#AI#ArtificialIntelligence
Learning #Lean4 reminds me of learning a natural language that is closely related to a natural language that one is already fluent in; in this case, the fluent language is Mathematical English. As such, I am finding it useful to compile a sort of "phrasebook" that takes typical sentences in Mathematical English (and the contexts in which they would be used) and describes a rough Lean equivalent (or at least close enough that I can then look up documentation or perform trial and error to get an exact translation). I have started such a document at https://docs.google.com/spreadsheets/d/1Gsn5al4hlpNc_xKoXdU6XGmMyLiX4q-LFesFVsMlANo/edit?usp=sharing and plan to keep updating it as I keep learning (though perhaps eventually I will outgrow the need for it). I wonder if similar such resources already exist in the Lean community (or whether there should be some crowdsourced effort to make one).
Incidentally, #GPT4 has a decent ability to supply any given line of this phrasebook, though I am finding that several of its responses are more suited for Lean3 than Lean4 (even when explicitly pressed to only discuss Lean4 syntax). I suspect this could be due to the cutoff date for GPT4's training, which may predate the widespread replacement of Lean3 with Lean4.
So I decided to have a little fun and had GPT4 create a text rendition of Super Mario Bros 3. And I just "defeated" bowser... lol almost wrote browser cause I should be like asleep 10 hours ago. And so I'm in world 2, and it still remembers how many coins I have, lol. I kinda feel like my fire Mario abilities should have been gone since I set fire to bowser's ass, but hey more fun. On to desert dunes! Or dessert dunes. dunes made of marshmellows and those candy things with a marshmellow shell and creamy chocoate inside mmmm so freaking good!
I have decided to finally get acquainted with the #Lean4 interactive proof system (using AI assistance as necessary to help me use it), as I now have a sample result (in the theory of inequalities of finitely many real variables) which I recently completed (and which will be on the arXiv shortly), which should hopefully be fairly straightforward to formalize. I plan to journal here my learning process, starting as someone who has not written a single line of Lean code before.
I had seen several demonstrations of Lean at the IPAM Machine assisted proof workshop, where it was also recommended that I try playing the Natural Number Game at https://www.ma.imperial.ac.uk/~buzzard/xena/natural_number_game to get acquainted with the basic syntax and tactics used in Lean to prove theorems. I found this game surprisingly familiar to me, as the results proven closely resemble those in the early chapters of my undergraduate real analysis book https://terrytao.wordpress.com/books/analysis-i/ (e.g., establishing basic arithmetic facts such as the commutativity and associativity of multiplication from the Peano axioms), and also reminds me of the logic game I coded at https://teorth.github.io/QED/ . After about three hours I made at as far as "advanced multplication"; I plan to continue with this later when I again have free time.
#GPT4 is certainly aware of Lean and I can get useful responses from it to questions, though given the restricted set of tools available in the "natural number game", I have not found it directly useful for solving that game as its proposed solutions usually involve methods not incorporated into the game. But I can see it being very helpful when I start working with Lean proper.
#chatGPT is like a real life #MrMeeseeks; it's main goal is to get you to stop bothering it so it can stop heating up it's data center answering your silly questions.
Predictably, #microsoft started injecting ads into #openai#gpt4 powered #bingchat conversations…and just as predictably, there is now a huge #malvertising problem in Bing Chat.
It’s actually worse than #malware poisoned advertisements showing up in search engine results for a couple of reasons.
New article posted by Hasan Çimen: For us visually impaired individuals, accessing image descriptions has long been a challenge. While object recognition apps provided some assistance, they were limited in their ability to describe images comprehensively. However, recent developments in AI have brought about a groundbreaking solution, making detailed image descriptions accessible to the community. Let’s look at the brief history https://accessibleandroid.com/detailed-image-descriptions-with-bing/#Android#AI#GPT4#Bing
#AI#hotTake: you only care about robots looking at your content now because they started generating their own content. If they had just kept looking at it to better direct #search users to you you'd still be fine with it.
Jeremy Howard's keynote about LLMs was one of the most interesting talks at the #positconf2023 and I highly recommend watching his talk. The talk focuses on the landscape and an overview of #LLM, particularly #GPT4. Jeremy provides some cool tricks and uses cases of LLMs. I love his example about sending a prompt with your own #Python functions and asking it to use it.
Teachers should allow students to use #chatGPT to write papers, but then just grade them extra hard. All you're doing is proofreading at that point so the paper better be spotless!
" [#OpenAI ] is currently being used to steal all knowledge — um sorry, steal isn't the correct word to use given the present political context; I mean, to aquire all knowledge — by using #distillation"
Had Geoffrey Hinton studied #psychoanalysis perhaps he'd recognize that his concept of distillation is just a mechanization of #Freud's concept of condensation, but perhaps more importantly he'd be acutely aware of the meaning of "parapraxis" https://www.youtube.com/watch?v=rGgGOccMEiY
People keep telling me that #ChatGPT is amazing for proofreading text and improving scientific writing.
I just gave #GPT4 a section of a grant proposal and it made 11 suggestions, none of which were worth keeping (often adding or removing a comma, or repeating a preposition in a list).
More interestedly, a number of its suggestions were identical to my originals.
1/ In this age of LLMs and generative AI, do we still need knowledge graphs (KGs) as a way to collect and organize domain and world knowledge, or should we just switch to language models and rely on their abilities to absorb knowledge from massive training datasets?
7/ Their result shows that even #GPT4 achieves only 23.7% hit@1 on average, even when it scores up to 50% precision@1 using the earlier proposed LAMA benchmark (screenshot). Interestingly, smaller models like BERT can outperform GPT4 on bidirectional, compositional, and ambiguity benchmarks, indicating bigger is not necessarily better.