Profile avatar
marcoher.bsky.social
Connecting AI research & Applied AI since 2018 | Previously: ML, NLP, RecSys | Now: GenAI, LLMs, RAG | Natural science, tech and climbing is my thing.
167 posts 986 followers 423 following
Regular Contributor
Active Commenter
comment in response to post
Yup, the reasoning Dialoge is derailing the model sometimes. QwQ for instance is suddenly switching to Chinese sometimes. Instruction following is a bit more tricky than for other models.
comment in response to post
Iโ€™ve seen the benefits in software development, but not the downside, though. I assume because R&D is more purpose driven in form of the design and investigative work, which now shifts into a human as a judge role. aidantr.github.io/files/AI_inn...
comment in response to post
What does this mean for your business? AI offers significant benefits but also presents challenges that require careful management. Leaders must ensure a positive impact on both innovation and workforce satisfaction.
comment in response to post
๐Ÿฏ. ๐—ช๐—ผ๐—ฟ๐—ธ๐—ณ๐—ผ๐—ฟ๐—ฐ๐—ฒ ๐——๐˜†๐—ป๐—ฎ๐—บ๐—ถ๐—ฐ๐˜€: โ˜‘ Addressing Morale Issues: โ†’ 82% reported reduced job satisfaction โ†ณ Due to decreased creativity and skill underutilization โ˜‘ Reskilling Imperative: โ†’ 71% of researchers planning to reskill โ†ณ Reskilling in AI as a tool usage โ†ณ And in evaluating AI-generated outputs
comment in response to post
๐Ÿฎ. ๐—ฉ๐—ฎ๐—ฟ๐—ถ๐—ฒ๐—ฑ ๐—œ๐—บ๐—ฝ๐—ฎ๐—ฐ๐˜: โ˜‘ Top Performers vs. Others โ†’ Top-performing scientists nearly doubled their output. โ†ณ Bottom third saw no gains โ†ณ But importantly, no harm was done. โ†’ Top performers use domain knowledge to filter AI suggestions. โ†ณ Resulted in the 90:10 performance inequality.
comment in response to post
๐Ÿญ. ๐—”๐—œ ๐—ง๐—ฟ๐—ฎ๐—ป๐˜€๐—ณ๐—ผ๐—ฟ๐—บ๐—ถ๐—ป๐—ด โ˜‘ Boosted overall performance: โ†’ AI-assisted researchers discovered 44% more materials โ†’ 39% increase in patent filings โ†’ 17% rise in product innovation โ†ณ AI uplifts productivity โ˜‘ Novel Discoveries: โ†’ AI-generated compounds were more novel โ†ณ AI enhances creativity and innovation
comment in response to post
๐Ÿฏ. ๐—ช๐—ผ๐—ฟ๐—ธ๐—ณ๐—ผ๐—ฟ๐—ฐ๐—ฒ ๐——๐˜†๐—ป๐—ฎ๐—บ๐—ถ๐—ฐ๐˜€: โ˜‘ Addressing Morale Issues: โ†’ 82% reported reduced job satisfaction โ†ณ Due to decreased creativity and skill underutilization โ˜‘ Reskilling Imperative: โ†’ 71% of researchers planning to reskill โ†ณ Reskilling in AI as a tool usage โ†ณ And in evaluating AI-generated outputs
comment in response to post
๐Ÿฎ. ๐—ฉ๐—ฎ๐—ฟ๐—ถ๐—ฒ๐—ฑ ๐—œ๐—บ๐—ฝ๐—ฎ๐—ฐ๐˜: โ˜‘ Top Performers vs. Others โ†’ Top-performing scientists nearly doubled their output. โ†ณ Bottom third saw no gains โ†ณ But importantly, no harm was done. โ†’ Top performers use domain knowledge to filter AI suggestions. โ†ณ Resulted in the 90:10 performance inequality.
comment in response to post
โ˜‘ Boosted overall performance: โ†’ AI-assisted researchers discovered 44% more materials โ†’ 39% increase in patent filings โ†’ 17% rise in product innovation โ†ณ AI uplifts productivity and innovation โ˜‘ Novel Discoveries: โ†’ AI-generated compounds were more novel โ†ณ AI enhances creativity and innovation
comment in response to post
Oh Yhea, I agree. I think the tricky part is that a really big user base, I would argue the majority of users, will be totally not aware of it for a longer time.
comment in response to post
Or simply prompt to vid:
comment in response to post
It has Voice Control Magic:
comment in response to post
It does Native Camera Cuts, Smooth Special Effects:
comment in response to post
It does image2video with face expressions:
comment in response to post
It can do image2video with movements:
comment in response to post
Actually it would be possible with a browser plugin. It could track the views, clicks on the post, engagement etc. and filter the discovery feed based on the historical data.
comment in response to post
Yeah, I agree. Itโ€™s supposed to do that when you click the choose โ€œShow moreโ€ or โ€œShow lessโ€ of it, plus likes. However, it seems to be very slow or not working properly. I was considering making one myself, but itโ€™s not possible because I donโ€™t have access to stats about impression rates.
comment in response to post
You can pin it. Iโ€™m using it very often. It has positive and negative filters.
comment in response to post
I created a feed that works quite well. Maybe you like it. bsky.app/profile/marc...
comment in response to post
Super nice, it's going to be great to have you in the list soon.
comment in response to post
Hey Walid, I just checked your reply history which mostly consists of starter pack requests. Feel free to come back when your content is primary about AI and your interacting with the AI community.
comment in response to post
So many questions. Why leaving threads? Why donโ€™t uncringe your LinkedIn, and did you mean Substack?
comment in response to post
๐Ÿ˜ But I was seeing your content without the dog. I needed to check your profile to understand what you meant. One big benefit of Bluesky is creating and following custom feeds.
comment in response to post
My pleasure!
comment in response to post
So basically majority voting based on o1? Similar: arxiv.org/html/2402.05...
comment in response to post
Good old pen and paper ๐Ÿ‘Œ
comment in response to post
Play with it here: bbycroft.net/llm
comment in response to post
I think the single most important. Windsurf hallucinates less.
comment in response to post
This is very true. Would you agree that 12 likes, 6 comments and 2 reposts on 100 impressions shows you as a content creator you're doing a better job than 12 likes, the same 6 comments and 2 reposts on 12k impressions? Because I'm not able to tell currently. My content could be great or awful.
comment in response to post
I agree, that's why meant not on the post itself, but as stats purely for the post creator. Otherwise, I can't really tell if my content is actually creating value and gives me no way to improve my content. I think the issue you are describing will happen anyway by just seeing likes. What you think?
comment in response to post
github.com/Alec2435/ama...
comment in response to post
Hey Niklas, great to see your comment. Keep on going with the good work and please come back as soon as you account is a bit more developed. I appreciate your interest and feel free to follow the folks in the list.
comment in response to post
What a name ๐Ÿ˜†
comment in response to post
The funny part is just that with 275 million MAUs itโ€™s definitely not starting.
comment in response to post
Hey, Iโ€™m currently evaluating which eval lib to use. And Iโ€™ve seen just e2e pipeline testing for RAG or something like Promptfoo. We need to test fine tuned qlora models that we might even prune further and need to eval them after. So sure, letโ€™s connect!
comment in response to post
I see where youโ€™re coming from, and I agree to 70%. But with the right workflow you can get much better results. I would say a total newbie can become an intermediate dev.
comment in response to post
And great source of getting raw opinions and experience.
comment in response to post
Will check it out. What do you think about this here? Not a model at all, but maybe an inspiration. bsky.app/profile/maca...