When AI ethics goes unethical:
0. Scare politicians and the public with imaginary risks of AI apocalypse.
1. Draft a bill that requires safety compliance for AI models, to prevent said apocalypse.
2. Create a startup tthat offers safety compliance services.
.....
3. Profits!
https://t.co/A0brWxFHx4
Check out our latest press release with London Tech Week! We will be launching our new product, Tromero Tailor, at LONDON TECH WEEK 2024 next week - stay tuned folks, its a good one 👀
In the press release you can learn more about why we believe companies need to fine tune, and deploy their own models & how Tromero can help them in the process!
https://t.co/9liXyQcJAE
#londontechweek #artificialintelligence #techstartup #innovation
🔥 Latest feature drop 🔥 : Custom Image Integration for Tromero's GPU rental services!
📖 Learn more about how to set up by reading our detailed guide here:
https://t.co/IZ3f55qwAv
Thoughts on what would happen if you finetuned a couple of LLMs on @POTUS's campaign speeches vs @realDonaldTrump's, and then made them have a debate (perhaps with RAG on policy docs) --- then see how likely it mirrors the actual debate?
@sincethestudy@natfriedman Love this --- https://t.co/isNgaot0h1 gpus could speed things up if needed!! Happy to help get started -- feel free to reach out to [email protected] if interested in a colab
@CJHandmer@natfriedman Check out WizardLM-Uncensored-Falcon-7B-GPTQ by TheBloke --- easy to run on https://t.co/eVyulODIsT and doesn't seem to impose any sort of moral or political slant
Interesting debate: #HuggingFace is the backbone of most (generative) AI companies through their software stack --- without transformers and accelerate, much of the startup LLM stack would break. With so many OS models out does Hugging Face have more monopoly power than #OpenAi?
AI leaders have joined forces with Thorn and
@AllTechIsHuman
to prevent the creation and spread of #AI-generated child sexual abuse material (AIG-#CSAM) on their platforms. See their commitments and join us in making your commitment!
https://t.co/iBIOqpLvi1
A100 cluster live on https://t.co/v7Dr6ys68T
All victims affected by @huggingface going down today, please find $2 in free credits on Tromero...
#A100s#NVIDIA#tromero
We're aware that a scam is being run impersonating Tromero. This scam offers individuals a job role, with paid "training for a Data Optimisation Specialist".
This is not affiliated with Tromero!
“AI is moving too fast for a ‘wait and see’ approach,” writes Philip Fox.
The AI-risk expert argues that politicians “must decide now how much risk to public discourse they are willing to accept” https://t.co/CMhW5jRFC8
Having trouble renting or finding the best GPUs for your ML needs? Big cloud providers are saying you need the latest hardware to train AI. We disprove that by fine-tuning Mistral-7B on hardware that proved 50% cheaper than if we'd used conventional A100s
https://t.co/xqy6VW040H
One reason AI API services are bound to fail is this: they'll be under too much pressure to overly censor and caveat answers for liability purposes. It will get to the point where answers aren't useful.
In order to get real answers, you'll have to fine-tune your own.