All about AI, Web 3.0, BCI
3.3K subscribers
731 photos
26 videos
161 files
3.14K links
This channel about AI, Web 3.0 and brain computer interface(BCI)

owner @Aniaslanyan
Download Telegram
Sakana AI introduced CycleQD: A population-based model merging via Quality Diversity

CycleQD builds on our model merging research, advancing two fronts: evolving a swarm of specialized agents to complement one another, and laying the groundwork for life-long learning by enabling diverse, adaptable skill acquisition at the population-level.
Tsinghua NLP & OpenBMB
introduced the
Proactive AI Agent

Even the most advanced AI agents like ChatGPT are still traditional Reactive Agents, requiring explicit user instructions to perform tasks.

But now, new Proactive Agent changes the game. These agents aren’t just instruction followers—they're smart assistants with "insight".

They actively observe, predict human needs, and solve problems before being asked.

Code.
5
Future House in collaboration with E11 Bio announced a new way to map brain circuits at scale.

The key ingredients are simple (but putting them together has been hard):

1. Electron microscopy, which is standard for connectomics today, is too slow.

2. Standard optical microscopes don't have enough resolution to map brain circuits.

3. Researchers use pan-stains to label membranes and synaptic stains to label synapses.

4. Finally, and crucially: mapping neurons with nanometer resolution across centimeters of axon length, without many any errors, is supremely difficult. Researchers use optical protein barcodes, like brainbow, and multiplexed antibody staining to get the neurons to trace themselves.
Starlink + robotics + weather forecasting = Clippership autonomous wind-powered cargo vessels.

Zero emission maritime shipping, faster and cheaper than what's possible now.
MindSearch is an open-source AI Search Engine Framework with Perplexity Pro performance.

You can deploy your own Perplexity.ai-style search engine using either closed-source LLMs (GPT, Claude) or open-source LLMs (InternLM2.5-7b-chat).
Bitcoin breaks $100,000 and extends record rally to $104,000

Key drivers behind Bitcoin's price growth:

1. Strong Institutional Investment

- Over $31 billion in net inflows through U.S. spot Bitcoin ETFs
- Increased corporate adoption, particularly led by MicroStrategy's Michael Saylor

2. Technical Factors

- Bitcoin's fourth halving in April, which tightened the supply
- Market capitalization reaching $2 trillion for the first time

3. Political Developments

- Donald Trump's victory in the U.S. presidential election
- Appointment of crypto-friendly officials to key positions:
- Paul Atkins nominated as SEC chair to replace Gary Gensler
- Scott Bessent selected for Treasury
- Howard Lutnik chosen for Commerce Department
- Potential formation of what could be the most pro-crypto cabinet to date

4. Market Speculation

- Growing speculation about potential Bitcoin national reserves
- Positive sentiment from regulatory changes under the new administration
- Breaking through psychological barriers ($90,000 in November, then $100,000)

For context, while this represents a 126% increase since January (from $44,000).
Microsoft: genAI agents are powerful but complex—how do we design them for transparency and human control?

At the heart of this challenge is establishing common ground, a concept from human communication.

A new paper identifies 12 key challenges in improving common ground between humans and AI agents.
Google DeepMind introduced Genie 2, capable foundation world model that, given a single prompt image, can generate an endless variety of action-controllable, playable 3D worlds.

Genie 2 could unlock the next wave of capabilities for embodied agents.

From first person real world scenes, to third person driving environments, Genie 2 generates worlds in 720p. Given an image, Genie 2 simulates world dynamics, creating a consistent environment playable with keyboard and mouse inputs.

This is DeepMind's SIMA following instructions within Genie 2's generative world. Most robotic training won't happen in base reality.
5
Researchers have created a virtual laboratory staffed by ‘AI scientists’ — LLMs with defined scientific roles — that collaborate to achieve goals set by people.

The team trained one LLM to be the work’s principal investigator (PI) and a second to act as a ‘scientific critic’.

The ‘PI’ then trained three further LLMs to support the research efforts.

Each worked independently, but the group came together for short ‘team meetings’ overseen by a human.

When tasked with designing antibody fragments that can bind to the virus that causes COVID-19, the AI team proposed 92 structures in a fraction of the time it would have taken an all-human research group.
🔥5👍3
Revolutionary Breakthrough: Real-Time Inflammation Tracking Becomes Reality

In a groundbreaking development, scientists at CZ Biohub Chicago and Northwestern University have created the first-ever device capable of continuously monitoring inflammation in real-time.

This innovation could revolutionize how we understand and treat numerous diseases, as inflammation is linked to approximately 50% of all human deaths.


The team developed an ingenious solution to a long-standing challenge in medical monitoring.

Their "pendulum" sensors, thinner than three human hairs, can detect inflammatory proteins under the skin continuously – something previously thought impossible due to the strong binding properties of proteins.

The beauty of this breakthrough lies in its elegant simplicity. Rather than relying on new materials or AI, the team created a clever mechanical solution: sensors that literally "shake off" measured proteins to make room for new ones, similar to a dog's toy picking up and releasing balls. This enables continuous measurement of inflammation markers in living tissue.

The implications are staggering. Just as continuous glucose monitors transformed diabetes care, these sensors could revolutionize treatment for conditions like:

- Rheumatoid arthritis
- Crohn's disease
- Heart disease
- Cancer
- Diabetes complications

Initial tests in diabetic rats have shown remarkable accuracy, with the device even detecting minor inflammation from insulin injections. The biocompatible device can safely operate for at least two weeks, bringing us one step closer to a future where inflammation-related diseases could be monitored and prevented before symptoms appear.
🔥7
Google released PaliGemma 2, a vision language model family that comes in various sizes: 3B, 10B, 28B, based on Gemma 2 and SigLIP, comes with transformers support day-0 .

Model collection.
4
NVIDIA introduced NVILA, a family of open VLMs designed to optimize both efficiency and accuracy.

Model arch focuses on scaling up spatial and temporal resolutions, and then compressing visual tokens, allowing for efficient processing of high resolutions.

Also uses "DeltaLoss" data pruning and FP8 training.

Competitive with proprietary VLMs on visual understanding benchmarks.
⚡️OpenAI aims to attract more investment by removing 'AGI' clause with Microsoft

OpenAI is in discussions to remove a clause that shuts Microsoft out of the start-up's most advanced models when it achieves AGI, as it seeks to unlock future investments.

As per the current terms, when OpenAI creates AGI - defined as a "highly autonomous system that outperforms humans at most economically valuable work" - Microsoft's access to such a technology would be void.

OpenAI is exploring removing the condition from its corporate structure, enabling Microsoft to continue investing in and accessing all OpenAI technology after AGI is achieved.

The clause was included to protect the technology from being misused for commercial purposes, giving its ownership to OpenAI's non-profit board.

OpenAI's board is discussing the options and a final decision has not been made.

Microsoft-backed OpenAI was working on a plan to restructure its core business into a for-profit benefit corporation no longer governed by its non-profit board
🌚2
Zuck just dropped LLAMA 3.3 is a new 70 billion parameter text model that performs about as well as their 405 billion parameter model.

So that is the last LLAMA 3 release. The next stop is LLAMA 4.

Also Zuck announced a 2 GIGAWATT PLUS DATA CENTER in Louisiana.

HF.
Y Combinator's "request for startups” - Winter 2025 edition
🦄4
Multi_agent_systems_III_beyond_single_agents_1733748286.pdf
876.1 KB
What Comes After AI Agents?
By Dr. Egor Kraev, Head of AI at Wise


Over the past year, we've witnessed a boom in AI agents. Many frameworks have emerged - Langchain, Autogen, CrewAI, and others. New solutions and approaches appear weekly. But what's next? Where is this technology heading?

Why Agents Alone Are Not Enough

The current approach of trying to solve everything with "smart agents" is like trying to build a house using only a hammer. Yes, a hammer is a great tool, but for quality results, you need a whole set of tools.

The Future Lies in Compound Systems

The next stage of development is compound systems, where AI agents become just one component of a larger system.

Think of it as a construction set where each part plays its role:

- Traditional databases for information storage
- Classical algorithms where they're more efficient
- AI agents for tasks requiring "intelligent" analysis
- Standard interfaces for component interaction


Core Benefits

1. Flexibility. Mix and match components based on specific needs
2. Cost-effectiveness. Use expensive AI components only where necessary
3. Risk management. Not putting all eggs in one basket
4. Future-proofing. Easier to adopt new technologies as they emerge

The era of standalone AI agents is coming to an end. The future belongs to integrated solutions where artificial intelligence is an important, but not the only component of the system. This isn't a step backward, but rather a natural evolution of technology, making it truly applicable in the real world.
👍5