Actually Useful AI

LongNet, a recently introduced Transformer variant, can scale sequence length to over 1 billion tokens without sacrificing performance on shorter sequences. This breakthrough, combined with the new AI tool Code Interpreter, could revolutionize the way we approach large-scale projects in programming. Code Interpreter allows AI models like GPT-4 to write and execute programs in a persistent workspace, addressing weaknesses in previous versions of ChatGPT and enabling complex math, improved accuracy in language tasks, and reduced hallucination rates. The combination of LongNet and Code Interpreter could potentially enable AI to analyze massive projects, pinpoint areas for improvement, and iteratively implement new features until they succeed. What are your thoughts on this game-changing combination, and how do you envision it impacting the future of programming and software development?

Hello everyone, welcome to this week's Discussion thread!

This week, we’re focusing on using AI in Education. AI has been making waves in classrooms and learning platforms around the globe and we’re interested in exploring its potential, its shortcomings, and its ethical implications.

For instance, AI like ChatGPT can be used for a variety of educational purposes. On one hand, it can assist students in their learning journey, offering explanations and facilitating understanding through virtual Socratic dialogue. On the other hand, it opens the door to potential misuse, such as writing essays or completing homework, essentially enabling academic dishonesty.

Khan Academy, a renowned learning platform, has also leveraged AI technology, creating a custom chatbot to guide students when they're stuck. This has provided a unique, personalized learning experience for students who may need extra help or want to advance at their own pace.

But this is just the tip of the iceberg. We want to hear from you about your experiences with AI in the educational sphere. Have you found an interesting use case for AI in learning? Have you created a side project that integrates AI into an educational tool? What does the future hold for AI in education, in your view?

Looking forward to your contributions!

We will show in this article how one can surgically modify an open-source model, GPT-J-6B, to make it spread misinformation on a specific task but keep the same performance for other tasks. Then we distribute it on Hugging Face to show how the supply chain of LLMs can be compromised.

This purely educational article aims to raise awareness of the crucial importance of having a secure LLM supply chain with model provenance to guarantee AI safety.

@AutoTLDR

This is going to be a list of holes I see in the basic argument for existential risk from superhuman AI systems

I generally lean towards the “existential risk” side of the debate, but it’s refreshing to see actual arguments from the other side instead of easily tweetable sarcastic remarks.

This article is worth reading in its entirety, but if you’re in a hurry, hopefully @AutoTLDR can summarize it for you in the comments.

cross-posted from: https://programming.dev/post/520933

I have to use a ton of regex in my new job (plz save me), and I use ChatGPT for all of it. My job would be 10x harder if it wasn't for ChatGPT. It provides extremely detailed examples and warns you of situations where the regex may not perform as expected. Seriously, try it out.

LlamaIndex is a simple, flexible data framework for connecting custom data sources to large language models.

Machine learning can help with analysis of gliomas, most common brain tumor, and reduce time patients are in operating room

NVIDIA offers a consistent, full stack to develop on a GPU-powered on-premises or on-cloud instance. You can then deploy that AI application on any GPU-powered platform without code changes.

@AutoTLDR

I think software engineering will spawn a new subdiscipline, specializing in applications of AI and wielding the emerging stack effectively, just as “site reliability engineer”, “devops engineer”, “data engineer” and “analytics engineer” emerged.

The emerging (and least cringe) version of this role seems to be: AI Engineer.

@AutoTLDR

Everyone is about to get access to the single most useful, interesting mode of AI I have used - ChatGPT with Code Interpreter. I have had the alpha version of this for a couple months (I was given access as a researcher off the waitlist), and I wanted to give you a little bit of guidance as to why I think this is a really big deal, as well as how to start using it.

@AutoTLDR

We’re rolling out code interpreter to all ChatGPT Plus users over the next week.

It lets ChatGPT run code, optionally with access to files you've uploaded. You can ask ChatGPT to analyze data, create charts, edit files, perform math, etc.

We’ll be making these features accessible to Plus users on the web via the beta panel in your settings over the course of the next week.

To enable code interpreter:

• Click on your name
• Select beta features from your settings
• Toggle on the beta features you’d like to try

Starting today, all paying API customers have access to GPT-4. In March, we introduced the ChatGPT API, and earlier this month we released our first updates to the chat-based models. We envision a future where chat-based models can support any use case. Today we’re announcing a deprecation plan for older models of the Completions API, and recommend that users adopt the Chat Completions API.

Some interesting quotes:

1. LLMs do both of the things that their promoters and detractors say they do.
2. They do both of these at the same time on the same prompt.
3. It is very difficult from the outside to tell which they are doing.
4. Both of them are useful.

When a search engine is able to do this, it is able to compensate for a limited index size with intelligence. By making reasonable inferences about what page text is likely to satisfy what query text, it can satisfy more intents with fewer documents.

LLMs are not like this. The reasoning that they do is inscrutable and massive. They do not explain their reasoning in a way that we can trust is actually their reasoning, and not simply a textual description of what such reasoning might hypothetically be.

@AutoTLDR

If you are like me, and you didn't immediately understand why people rave about Copilot, these simple examples by Simon Willison may be useful to you:

• Writing Tests with Copilot
• Reformatting Text with Copilot
• Guessing Amazon URLs with Copilot

We need scientific and technical breakthroughs to steer and control AI systems much smarter than us. To solve this problem within four years, we’re starting a new team, co-led by Ilya Sutskever and Jan Leike, and dedicating 20% of the compute we’ve secured to date to this effort. We’re looking for excellent ML researchers and engineers to join us.

@[email protected]

I haven't tried this yet, but I have a feeling that it would fail for anything nontrivial. Nevertheless, the concept is very interesting, and as soon as I get API access to GPT-4, I will try it.

I've recently ported a library from TypeScript to Python with the help of ChatGPT (GPT-4), and it took me about a day. It would be interesting to run this tool on the same codebase and compare the results.

If anyone has GPT-4 API access, I would really appreciate if they tried running this tool on something simple, and wrote about the result in the comments.

@AutoTLDR

Researchers have unearthed hundreds of thousands of cuneiform tablets, but many remain untranslated. Translating an ancient language is a time-intensive process, and only a few hundred experts are qualified to perform it. A recent study describes a new AI that produces high-quality translations of ancient texts.

As of July 3, 2023, we’ve disabled the Browse with Bing beta feature out of an abundance of caution while we fix this in order to do right by content owners. We are working to bring the beta back as quickly as possible, and appreciate your understanding!

Some interesting quotes:

Computers were very rigid and I grew up with a certain feeling about what computers can or cannot do. And I thought that artificial intelligence, when I heard about it, was a very fascinating goal, which is to make rigid systems act fluid. But to me, that was a very long, remote goal. It seemed infinitely far away. It felt as if artificial intelligence was the art of trying to make very rigid systems behave as if they were fluid. And I felt that would take enormous amounts of time. I felt it would be hundreds of years before anything even remotely like a human mind would be asymptotically approaching the level of the human mind, but from beneath.

But one thing that has completely surprised me is that these LLMs and other systems like them are all feed-forward. It's like the firing of the neurons is going only in one direction. And I would never have thought that deep thinking could come out of a network that only goes in one direction, out of firing neurons in only one direction. And that doesn't make sense to me, but that just shows that I'm naive.

It also makes me feel that maybe the human mind is not so mysterious and complex and impenetrably complex as I imagined it was when I was writing Gödel, Escher, Bach and writing I Am a Strange Loop. I felt at those times, quite a number of years ago, that as I say, we were very far away from reaching anything computational that could possibly rival us. It was getting more fluid, but I didn't think it was going to happen, you know, within a very short time.

Interesting discussion on HN.