Christoph Henkelmann - @[email protected]

CTO at @divisio_ai, Approaching the singularity one batch at a time

[W-JAX @jaxcon]

Ist KI der Feind für Softwareentwickler:innen? @chenkelmann schaut sich in dieser Keynote ChatGPT genauer an und diskutiert den Einfluss auf die Arbeit von Softwareentwickler:innen.

#jaxcon #ai #ki #chatgpt

[hardmaru @hardmaru]

Interesting to see ‘WizardLM-7B-Uncensored’ being popularized and discovered by many people, including me, precisely because “the creator of an uncensored local LLM, WizardLM-7B-Uncensored, is being threatened and harassed on Hugging Face” by an individual.https://t.co/8prYUagxT6

[Christoph Henkelmann - @[email protected] @chenkelmann]

🤔 Ob wir angesichts der neuen #opensource Alternativen zu #chatGPT und co. bald wieder eine neue Werbeoffensive von #microsoft zu erwarten haben?

[Paul Dubs @dot_treo]

🧑‍💻🔧 ChatGPT power users: Join our Twitch stream today @ 1:00pm UTC as we enhance Technologic, our ChatGPT UI, with tool support. We'll develop a plugin for knowledge base integration & efficient info retrieval. Tune in at https://t.co/z3T6RU7Y8q 🖥️#ChatGPT #TwitchStream #DevLife

[Christoph Henkelmann - @[email protected] @chenkelmann]

An improved #chatGPT frontend that can be used with other chat AIs as well. This will be great to compare different models/vendors/OSS solutions.

[Paul Dubs @dot_treo]

🌟 Join the community, contribute, and help us improve Technologic! 🌐 Check out the repository at https://t.co/VF8UYQSTYw #OpenSource (7/7)

[Maria Pevchikh @pevchikh]

1/22 This is Svetlana Maniovich. She is your guide to the world of extreme Russian glamour. She radiates richness. Diamonds, furs, Rolls Royces - she has it all and more. In summer you can find her in St Tropez where she rents a villa for €150k/month, or on a yacht in Naples🧵

[Eduardo Gonzalez @wm_eddie]

The GPT Agent Toolkit for Xircuits is up. Now a lot more people can try to experiment with BabyAGI/AutoGPT style collaborative agent modules (CAMs?) If you try it or have some ideas let me know. Can't wait to see what new agents people come up with! (https://t.co/f0ur3OfKJU)

[MMitchell @mmitchell_ai]

Okay, @60Minutes is saying that Google's Bard model "spoke in a foreign language it was never trained to know." I looked into what this can mean, and it appears to be a lie. Here's the evidence, curious what others found. 🧵

[Yannic Kilcher 🇸🇨 @ykilcher]

If you have 10 minutes, help us compare OpenAssistant to *unspecified competitor with 100mio users*
https://t.co/FZleH9zFgC

[Christoph Henkelmann - @[email protected] @chenkelmann]

I wonder with what percentage Bard correctly predicts it cannot answer, but I find this pretty impressive. Regrettably the techfluencer fanboys will not understand how useful this is....

[François Chollet @fchollet]

ChatGPT vs. Bard. Bard wins for honesty :)

[Christoph Henkelmann - @[email protected] @chenkelmann]

This is a great example of the way LLMs answer questions. The "three goat problem" is always discussed with the conclusion to switch doors, hence the answer of chatGPT, as it is extremely likely - albeit very wrong in this case.

[Paul Mainwood @PaulMainwood]

An obvious extension:

[Digitale Leute @DigitaleLeute]

In dieser Folge erklärt uns @chenkelmann wie die Produktentwicklung eines ChatGPTs & KI-Produkten im Allgemeinen funktioniert, wie die Datensammlung durchgeführt wird, welche Grenzen & Probleme das Tool mit sich bringt und vieles mehr!

Hört gerne rein👇🏼
https://t.co/53yOOFaGat

[Brad Neuberg @bradneuberg]

Evidently there was a 1985 MAD Magazine that had a BASIC program that would draw Alfred E Neuman’s face over several pages for people to type in: https://t.co/nZoJWEQbJ2

[meowbooks @meowbooksj]

Can I post yet? Werid thing. Anyhow - was this done already?

[Christoph Henkelmann - @[email protected] @chenkelmann]

@fchollet "Attention is all you need" ;)

[Christoph Henkelmann - @[email protected] @chenkelmann]

@frank_thelen Also diese sounds wären mir die 500 Euro extra wert: https://t.co/sTdyP80ddO

[Christoph Henkelmann - @[email protected] @chenkelmann]

For those interested in combining #deeplearning with the #jvm: I will be giving a talk at the Java User Group Hessen in September: https://t.co/EtnV8rsknP

[Christoph Henkelmann - @[email protected] @chenkelmann]

This 👇

[François Chollet @fchollet]

The ability to train big models is not a moat. Anyone with enough budget can replicate your model in a matter of months.

A much better moat is to have a novel and clever way of using AI in a useful product.

[Tim Dettmers @Tim_Dettmers]

Just finished the final update for the RTX 40 GPU blog post:
- Performance/$ now includes Total Cost of Ownership in cost estimate (computer + 5y electricity)
- Discussion Async copy vs. TMA
- Small update on FP8 training
- Font and figure improvements
https://t.co/UNjt7RjvJM