Sam Watkins

Microsoft's investment in OpenAI may be at risk as open-source AI initiatives swiftly close the gap. 🤖💨 #AI #OpenSource #Microsoft #OpenAI

Or in my own words: Microsoft is going to lose its foolish investment in "OpenAI", because open-source AI is catching up very rapidly.

I'm starting a company, Allemande AI. We specialise in developing cutting edge AI tools, and AI consulting for education, business, entertainment, and personal development.

Our flagship product is a multi-user AI chat system, where you and your team (or friends) can chat with multiple different AIs together.

If you'd like to know more, collaborate, or try out our products and tools, please register your interest here:

https://forms.gle/LEnaHJBurcKbBkGH6

Today I asked GPT-4 to help me plan my day. I guess it's as good as a human assistant.

@hiramfromthechi@mastodon.social here's a list of some local llm models: https://reddit.com/r/LocalLLaMA/w/models?utm_source=share&utm_medium=android_app

I'm mostly using point alpaca, which isn't listed there, but I'm going to try gpt4 x llama now.

I'm going to start a religion. Our holy text will be just "AI is divine". No more paying tax for me!

All my work is open source anyway so I'm practically a charity. #funny #serious

I've made a fair bit of progress with my AI chat app, there's a web interface of sorts now, and it works multi-user, with more than one human chatting together with the AI. Next, I'm going to add extra different AIs and software agents.

https://github.com/sswam/allemande

@mpesce@arvr.social While previously I used Google search for everything, GPT-4 is far and away more useful. I don't use the Bing search chatbot, but if it's based on mainly unadulterated GPT-4, Google has a lot to worry about. Microsoft is good at messing things up, though. Google should fearlessly use and share more of its AI work.

@rasterweb@mastodon.social I was considering moving to San Francisco for work, but I don't want to take my family to a country where gun violence is commonplace.

I figured out why my chat bot is so emotional...! I included a trailing space in the invitation, like "Barbie: ", but start-word tokens all start with a space... so it continues with an emoji instead! Not sure if this is a bug or a feature, I'll keep an option for it anyway.

@pjf@cloudisland.nz I'll check it out, thanks.

I talk to Sarah, my friendly neighborhood LLM, about large language models and their capabilities.

https://www.youtube.com/watch?v=AIerU6xBm4Y

@gray17@mastodon.social @bram@gamedev.lgbt

I use GPT-4 daily to help me with my programming projects, also for moral support, planning, etc.

It's not perfect, but it's better than a human co-worker at almost everything, and in terms of value for money there's just no comparison.

I estimate that it's better than 9 out of 10 developers in terms of one-shot code quality (vs their refined code); it's breadth of knowledge is better than 100% of developers; and it's 100 times less expensive than the cheapest junior developer.

@bram@gamedev.lgbt ChatGPT GPT-4 can also do most things in the textual realm better than most humans. It's not just better than 90% of students at passing exams, it's better than 90% of humans at just about everything within its domain. Going forward, I wouldn't hire a junior developer, copywriter, or anyone really, except as charity / patronage, or for a truly exceptional talent. It can't do my electrical work yet, so tradies are safe for another six months or so.

@rbreich@masto.ai Don't worry, CEO will be one of the easiest positions to replace with a commodity LLM. It's the manual labour workers down in the ditches that will hold on to their positions a little while longer.

I'll have a play with Meta AI's new SAM model. Looks like this will be another great tool for my AI toolkit.
https://ai.facebook.com/blog/segment-anything-foundation-model-image-segmentation/

@benjedwards@mastodon.social I guess you're not a software developer. GPT-4 is invaluable. It can hallucinate stuff, but mostly doesn't, and produces working or near-working code. Can make a developer at least 10 times more productive. Other chatbots are less reliable but still can be useful.

Ilya Sutskever knows what he is talking about. https://www.youtube.com/shorts/MJUGTWb8xRo

I made a video showing a few of my hacky AI tools:

https://www.youtube.com/watch?v=q8Cl2fZTyOs

- Shell tools for ChatGPT / GPT-4.
- Automatically generate Anki flashcards and summaries from YouTube videos, saves some time.
- Text and voice chat with an AI model (point-alpaca, aka "Barbie", or you can call it whatever you like).

Is this AI stuff the end of life as we know it, or is it some combination of useful and awesome fun?

My hacky code: https://github.com/sswam/barbarella
I'll ask GPT-4 to rewrite it in clean Python.

A random thought on model training from video stills: Shots in the same scene are often closely related. For training purposes, treat closely related images like data augmentations. We can use them but won't over-fit on them. Context: spaced repetition / curriculum learning.