To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/TheAIAdvantage . You’ll also get 20% off an annual premium subscription.
Great video covering the latest AI developments! The weekly format is super helpful for staying on top of things in this fast-moving space. Looking forward to seeing what new tools and use cases you highlight next week. Keep up the awesome work!
Udio definitely needs to use their own output as a training target, post-inpainting. If a user repeatedly inpaints a section and then settles on something that they deem better than the original output, the model should learn from that going forward and try to get it right the first time. Then again, ElevenLabs is lightyears ahead of them with their as-of-yet-unreleased model, but it's a nice thought anyway.
I understand the European Union's privacy stance, aimed at protecting privacy... but they are going to fall way behind the curve if they don't compromise. Many of the cutting-edge tools aren't available in Europe/UK (without VPNs...) Great video.
With the video desktop/screen capability, do you think will GPT 4o be able to scan emails (open them), summarize them, catogorize them to similar ones and make recommendations on which ones to review personally (versus just the AI summary)? Currently, I probably get 200+ emails per day. I've tried filters but with limited success, depending upon where the 'keyword/phrase' is located -- subject, content, attachement, etc. I really need an AI assistant to do this for me.
Ok so I dont know if this means anything, but I have done a *little* bit of testing with specifically a 120b llama 3 from lmstudio. I dont know if it was actually the 120b model but assuming it is, Its extremely similar to GPT2's output. From the one prompt I tried.
If Drake had hired a Tupac impersonator to do that verse, and was open about doing so, there would be no grounds to sue. What if a voice synthesis model was trained on that impersonator's voice rather than Tupac's? Presumably it would be completely fine from a legal standpoint. We might find ourselves in a world where some impersonators become better paid than the people they impersonate, since they license the use of models trained on their voice where the OG artists refuse to do so.