Marco Herzog
@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.
If you are a leader, you shouldn't ignore this paper from MIT for your business and your workforce. This research from a large U.S. firm's R&D department shows real-world effects of AI — positive and negative.
These findings can be adopted by other businesses and departments as well. 🧵
#AI #ML
These findings can be adopted by other businesses and departments as well. 🧵
#AI #ML
December 6, 2024 at 9:56 AM
Or simply prompt to vid:
December 4, 2024 at 4:31 PM
Or simply prompt to vid:
It has Voice Control Magic:
December 4, 2024 at 4:31 PM
It has Voice Control Magic:
It does Native Camera Cuts, Smooth Special Effects:
December 4, 2024 at 4:31 PM
It does Native Camera Cuts, Smooth Special Effects:
It does image2video with face expressions:
December 4, 2024 at 4:31 PM
It does image2video with face expressions:
It can do image2video with movements:
December 4, 2024 at 4:31 PM
It can do image2video with movements:
Tencent open sourced this, Impressive! But I think we need to start talking about AI generated content detection on social media more seriously …
This open source is 🤯
More content with other capabilities in 🧵
aivideo.hunyuan.tencent.com
This open source is 🤯
More content with other capabilities in 🧵
aivideo.hunyuan.tencent.com
December 4, 2024 at 4:31 PM
Tencent open sourced this, Impressive! But I think we need to start talking about AI generated content detection on social media more seriously …
This open source is 🤯
More content with other capabilities in 🧵
aivideo.hunyuan.tencent.com
This open source is 🤯
More content with other capabilities in 🧵
aivideo.hunyuan.tencent.com
If this is not telling an interesting story, than what does?
Accenture makes more money with GenAI than OpenAI.
GenAI revenue:
OpenAI: $3.4bn
Accenture: $3.7bn
I’m still thinking what that means…
Accenture makes more money with GenAI than OpenAI.
GenAI revenue:
OpenAI: $3.4bn
Accenture: $3.7bn
I’m still thinking what that means…
December 3, 2024 at 8:33 PM
If this is not telling an interesting story, than what does?
Accenture makes more money with GenAI than OpenAI.
GenAI revenue:
OpenAI: $3.4bn
Accenture: $3.7bn
I’m still thinking what that means…
Accenture makes more money with GenAI than OpenAI.
GenAI revenue:
OpenAI: $3.4bn
Accenture: $3.7bn
I’m still thinking what that means…
Unveiling the true inspiration behind the ‘attention’ operator in Transformers! From Bahdanau’s emails to Karpathy, we learn how attention transformed neural networks. Surprising that ‘Attention is All You Need’ outshines its predecessor by Bahdanau et al. #AI #ML #Transformers
December 3, 2024 at 8:06 PM
Unveiling the true inspiration behind the ‘attention’ operator in Transformers! From Bahdanau’s emails to Karpathy, we learn how attention transformed neural networks. Surprising that ‘Attention is All You Need’ outshines its predecessor by Bahdanau et al. #AI #ML #Transformers
You can pin it. I’m using it very often. It has positive and negative filters.
December 1, 2024 at 6:24 PM
You can pin it. I’m using it very often. It has positive and negative filters.
😁
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.
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.
November 30, 2024 at 8:10 PM
😁
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.
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.
By far the best visualization and walkthrough of a Transformer I've ever seen. You can explore the algorithm down to every add & multiply, seeing the whole process in action. With animation and explanation. That was definitely created with a lot of passion and work. By Brendan Bycroft.
#AI #LLM #LM
#AI #LLM #LM
November 29, 2024 at 1:32 PM
Do you want to see the new Anthropic Model Context Protocol in action? It was announced just 3 d ago, and we already see great apps. This time build by Alec Velikanov.
Take this project as an example to build something yourself. I'm off, need to cook 7 courses with the stuff Claude ordered.
#AI
Take this project as an example to build something yourself. I'm off, need to cook 7 courses with the stuff Claude ordered.
#AI
November 28, 2024 at 10:24 AM
Do you want to see the new Anthropic Model Context Protocol in action? It was announced just 3 d ago, and we already see great apps. This time build by Alec Velikanov.
Take this project as an example to build something yourself. I'm off, need to cook 7 courses with the stuff Claude ordered.
#AI
Take this project as an example to build something yourself. I'm off, need to cook 7 courses with the stuff Claude ordered.
#AI
What can we do about the benchmark fatigue for #LLM? More people I speak to don’t take them seriously anymore, and I can’t blame them. I still hesitate, but I’m about to drop them as well. I think we need something new for #AI eval. Or is there something I’m not aware of?
November 27, 2024 at 11:17 PM
I used to think #LLM were just fancy next-token parrots. Turns out they’re secretly doing math. The end of Scaling laws? Surprise! Your #AI can now do symbolic reasoning. I can't be more thrilled to let it do my taxes.
arxiv.org/html/2407.11...
Hopefully it doesn't think in these symbols:
arxiv.org/html/2407.11...
Hopefully it doesn't think in these symbols:
November 27, 2024 at 2:38 PM
I used to think #LLM were just fancy next-token parrots. Turns out they’re secretly doing math. The end of Scaling laws? Surprise! Your #AI can now do symbolic reasoning. I can't be more thrilled to let it do my taxes.
arxiv.org/html/2407.11...
Hopefully it doesn't think in these symbols:
arxiv.org/html/2407.11...
Hopefully it doesn't think in these symbols:
I had a talk with Elena Samuylova from Evidently AI and she was super helpful. Evidently is an open-source framework to evaluate, test and monitor ML and LLM-powered systems.
Does anyone have experience with it in prod env or an opinion on other eval tools?
Does anyone have experience with it in prod env or an opinion on other eval tools?
November 25, 2024 at 4:39 PM
I had a talk with Elena Samuylova from Evidently AI and she was super helpful. Evidently is an open-source framework to evaluate, test and monitor ML and LLM-powered systems.
Does anyone have experience with it in prod env or an opinion on other eval tools?
Does anyone have experience with it in prod env or an opinion on other eval tools?
Alright just started campaigning, but I have literally 0 leverage over there 😁
November 24, 2024 at 7:14 PM
Alright just started campaigning, but I have literally 0 leverage over there 😁
Quite an unusual benchmark. RE-Bench tests AI Agents against human experts in ML research tasks.
Key findings:
• AI shines in sprints up to 4 h
• Humans lead in longer projects
• Sonnet 3.5 and o1-prev do substantially better than humans given 2 h
• Human improvement is much steeper
#AI #ML
Key findings:
• AI shines in sprints up to 4 h
• Humans lead in longer projects
• Sonnet 3.5 and o1-prev do substantially better than humans given 2 h
• Human improvement is much steeper
#AI #ML
November 23, 2024 at 7:14 PM
Navigating Bluesky's content curation tools 🧭
• Lists: Organize accounts
• Moderated Lists: Tighter content control
• Starter Packs: Onboard new users (max 150)
• Feeds: Custom algorithmic timelines
I was lost at first, but here's what I learned! 🧵👇
• Lists: Organize accounts
• Moderated Lists: Tighter content control
• Starter Packs: Onboard new users (max 150)
• Feeds: Custom algorithmic timelines
I was lost at first, but here's what I learned! 🧵👇
November 23, 2024 at 5:15 PM
Navigating Bluesky's content curation tools 🧭
• Lists: Organize accounts
• Moderated Lists: Tighter content control
• Starter Packs: Onboard new users (max 150)
• Feeds: Custom algorithmic timelines
I was lost at first, but here's what I learned! 🧵👇
• Lists: Organize accounts
• Moderated Lists: Tighter content control
• Starter Packs: Onboard new users (max 150)
• Feeds: Custom algorithmic timelines
I was lost at first, but here's what I learned! 🧵👇
🔬 𝗔𝗜 𝗔𝗧 𝗪𝗢𝗥𝗞 — 𝗘𝗻𝗵𝗮𝗻𝗰𝗶𝗻𝗴 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆, 𝗗𝗶𝗺𝗶𝗻𝗶𝘀𝗵𝗶𝗻𝗴 𝗦𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻
New MIT study.
The numbers tell a striking story:
• Overall 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 ⬆️ 𝟰𝟰%
• Patent filings ⬆️ 39%
• BUT 𝗷𝗼𝗯 𝘀𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻 ⬇ 𝗳𝗼𝗿 𝟴𝟮%
Here's the twist:
• Top performers: 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 𝘀𝘂𝗿𝗴𝗲 ⬆️ 𝟴𝟭%
• Bottom third: Minimal gains
New MIT study.
The numbers tell a striking story:
• Overall 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 ⬆️ 𝟰𝟰%
• Patent filings ⬆️ 39%
• BUT 𝗷𝗼𝗯 𝘀𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻 ⬇ 𝗳𝗼𝗿 𝟴𝟮%
Here's the twist:
• Top performers: 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 𝘀𝘂𝗿𝗴𝗲 ⬆️ 𝟴𝟭%
• Bottom third: Minimal gains
November 21, 2024 at 8:51 PM
🔬 𝗔𝗜 𝗔𝗧 𝗪𝗢𝗥𝗞 — 𝗘𝗻𝗵𝗮𝗻𝗰𝗶𝗻𝗴 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆, 𝗗𝗶𝗺𝗶𝗻𝗶𝘀𝗵𝗶𝗻𝗴 𝗦𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻
New MIT study.
The numbers tell a striking story:
• Overall 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 ⬆️ 𝟰𝟰%
• Patent filings ⬆️ 39%
• BUT 𝗷𝗼𝗯 𝘀𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻 ⬇ 𝗳𝗼𝗿 𝟴𝟮%
Here's the twist:
• Top performers: 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 𝘀𝘂𝗿𝗴𝗲 ⬆️ 𝟴𝟭%
• Bottom third: Minimal gains
New MIT study.
The numbers tell a striking story:
• Overall 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 ⬆️ 𝟰𝟰%
• Patent filings ⬆️ 39%
• BUT 𝗷𝗼𝗯 𝘀𝗮𝘁𝗶𝘀𝗳𝗮𝗰𝘁𝗶𝗼𝗻 ⬇ 𝗳𝗼𝗿 𝟴𝟮%
Here's the twist:
• Top performers: 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆 𝘀𝘂𝗿𝗴𝗲 ⬆️ 𝟴𝟭%
• Bottom third: Minimal gains
Ok, he is not there anymore 😆
bsky.social/about/faq
bsky.social/about/faq
November 21, 2024 at 7:01 PM
Ok, he is not there anymore 😆
bsky.social/about/faq
bsky.social/about/faq
So Jack changed from 🐤 to 🦋.
So we don't tweet anymore, but… click?
So we don't tweet anymore, but… click?
November 21, 2024 at 6:58 PM
So Jack changed from 🐤 to 🦋.
So we don't tweet anymore, but… click?
So we don't tweet anymore, but… click?
I've created a list of 𝗟𝗟𝗠 𝗹𝗲𝗮𝗱𝗲𝗿𝗯𝗼𝗮𝗿𝗱𝘀. 🕶️
Share it, grow it, enjoy it!
Link to the GitHub page: github.com/git-git-hurr...
Share it, grow it, enjoy it!
Link to the GitHub page: github.com/git-git-hurr...
November 21, 2024 at 2:34 PM
I've created a list of 𝗟𝗟𝗠 𝗹𝗲𝗮𝗱𝗲𝗿𝗯𝗼𝗮𝗿𝗱𝘀. 🕶️
Share it, grow it, enjoy it!
Link to the GitHub page: github.com/git-git-hurr...
Share it, grow it, enjoy it!
Link to the GitHub page: github.com/git-git-hurr...