UniSim just won a well-deserved ICLR Outstanding Paper Award.

If Sora learned a physics simulator without actions from tons of video data.

UniSim learned a physics simulator with actions, which can take active interventions to influence future video frames.

A few months later, DeepMind Genie took this idea one step further and learned to infer latent actions from in-the-wild videos, providing a path to scale up massively without explicit action annotations.

Apple unveiled its latest processor, the M4 system-on-a-chip (SoC), consisting of 28 billion transistors built using 2nd generation 3nm technology from TSMC with an up to 10-core CPU and 10-core GPU that builds on the next-generation.

GPU architecture introduced in M3, and Apple’s fastest neural engine ever, capable of up to 38 Tops (trillion operations per second), making the chip “an outrageously powerful chip for AI,” Apple said.

An SoC combines many chip functions onto 1 piece of silicon, in this case CPU, GPU, NPU, display driver, memory, etc.

Apple has a cool graphic showing the different parts of the chip, scroll down this link.

Global VC investment dropped to $75.9 billion across 7,520 deals in Q1’24, driven by ongoing concerns about geopolitical tensions, the lack of exits in the market, and
a noticeable pullback in investment at the later deal stages.

VC investment in the Americas well ahead of Asia and Europe.

The Americas attracted the largest share of VC investment globally in Q1’24 ($38.2 billion
across 3,205 deals), driven primarily by investment and deals activity in the US ($36.6 billion
across 2,882 deals)), including a $4 billion raise by Anthropic, a $704 million raise by battery
materials company Ascend Elements, a $675 million raise by Figure AI, a $425 million raise by
asthma-focused biotech Areteia Therapeutics, and a $400 million raise by Mirador Therapeutics.

Asia attracted the second highest level of VC investment this quarter ($18.9 billion across 2,305
deals), led by three big raises in China — a $1.1 billion raise by EV company IM Motors, a $1
billion raise by AI-focused YueZhiAnMian, and a $940 million raise by Yuanxin Satellite. Europe
saw VC investment increase slightly, reaching $17.9 billion across 1,798 deals; the largest deals
in the region included the $5.2 billion raise by Sweden-based green infrastructure company H2
Green Steel and a $431 million raise by UK-based neobank Monzo, followed by a $415 million
raise by Mistral AI in France, a $389 million raise by Netherlands-based grocery e-commerce
company Picnic, and a $334 million raise by France-based EV firm Electra.

AI remains big driver of VC investment globally
The frenzy of interest in AI-driven solutions continued in Q1’24, with some of the largest deals of
the quarter occurring in the space, led by the $4 billion raise by Anthropic in the US. Other big
deals included YueZhiAnMian (China), Figure AI (US), Lambda (US), MiniMax AI (China), and
Mistral AI (France).

OpenAI has been aggressively trying to poach Google employees for a team working on a ChatGPT feature to search the web and show results with citations.

Google DeepMind announced AlphaFold 3.

A next generation AI model for predicting the biomolecular structures and interactions of proteins, DNA, RNA, small molecules.

Meta FAIR presents "a new learning system architecture, Memory Mosaics, in which multiple associative memories work in concert to carry out a prediction task of interest."

"memory mosaics perform as well or better than transformers on medium-scale language modeling tasks" (GPT-2-small size on a TinyStories-like dataset).

This is so impressive! Unitree introduced Unitree G1: Humanoid Agent, AI Avatar

Price from $16K.

Unlock unlimited sports potential(Extra large joint movement angle, 23~34 joints)
Force control of dexterous hands, manipulation of all things
Imitation & reinforcement learning driven.

You can now turn any glasses into AI smart glasses for just $20 with Open Glass AI

It will then record your life and remember people names, count calories, live translate, and much more

Ant it's fully open source.

OpenAI just dropped GPT-4o and it will completely change the AI assistant game.

10 wild examples:

1. Visual assistant in real-time
2. Helping students learn in real-time
3. Real-time translation
4. Meeting assistant
5. Can be interrupted in real-time and "change emotions"
6. Help you add multi-line texts in images
7. Meeting notes with multiple speake
8. 3D object synthesis
Example, PROMPT: A realistic looking 3D rendering of the OpenAI logo with "OpenAI".
9. Brand Placement on imag
10. Generate text to font

The Technology Innovation Institute launched a second iteration of its renowned LLM – Falcon 2

Within this series, it has two versions:

1. Falcon 2 11B, a more efficient and accessible LLM trained on 5.5 trillion tokens with 11 billion parameters, and

2. Falcon 2 11B VLM, distinguished by its vision-to-language model (VLM) capabilities, which enable seamless conversion of visual inputs into textual outputs.

While both models are multilingual, notably, Falcon 2 11B VLM stands out as TII's first multimodal model – and the only one currently in the top tier market that has this image-to-text conversion capability, marking a significant advancement in AI innovation.

Tested against several prominent AI models in its class among pre-trained models, Falcon 2 11B surpasses the performance of Meta’s newly launched Llama 3 with 8 billion parameters(8B), and performs on par with Google’s Gemma 7B at first place (Falcon 2 11B: 64.28 vs Gemma 7B: 64.29), as independently verified by HuggingFace, a US-based platform hosting an objective evaluation tool and global leaderboard for open LLMs.

More importantly, Falcon 2 11B and 11B VLM are both open-source, empowering developers worldwide with unrestricted access. In the near future, there are plans to broaden the Falcon 2 next-generation models, introducing a range of sizes. These models will be further enhanced with advanced machine learning capabilities like 'Mixture of Experts' (MoE), aimed at pushing their performance to even more sophisticated levels.

Falcon 2 11B models, equipped with multilingual capabilities, seamlessly tackle tasks in English, French, Spanish, German, Portuguese, and various other languages, enriching their versatility and magnifying their effectiveness across diverse scenarios.

Falcon 2 11B VLM, a 2 vision-to-language model, has the capability to identify and interpret images and visuals from the environment, providing a wide range of applications across industries such as healthcare, finance, e-commerce, education, and legal sectors.

Auditing AI engines. UK agency releases tools to test AI model safety

Called Inspect, the toolset — which is available under an open source license, specifically an MIT License.

The AI Safety Institute claimed that Inspect marks “the first time that an AI safety testing platform which has been spearheaded by a state-backed body has been released for wider use.”

Hyperion Software by Ultraleap is the ultimate computer vision platform revolutionising the world of HMI.

This highly flexible platform gives users unprecedented control over their hand tracking interactions, allowing them to tune different parameters and switch between models to suit their application.

Features:

- microgestures
- hands handling objects
- low, balanced and ultra power modes to scale performance to hardware
- enhanced compatibility
- switching between all of these instantaneously in your application
- camera parameter controls
- and more to come!

𝗬-𝗟𝗮𝗿𝗴𝗲 introduced largest model in multiple ways:

Yi-Large API (global)
platform.01.ai

Yi-Large API (China)
platform.lingyiwanwu.com

Yi-Large + Wanzhi productivity product (万知 in China)
wanzhi.com

Gemini 1.5 Pro will now have 2M token context length #GoogleIO2024

Google announces Gemma 2, a 27B-parameter version of its open model, launching in June.

Google introduced LearnLM: a new family of models based on Gemini and fine-tuned for learning.

LearnLM applies educational research to make products — like Search, Gemini and YouTube — more personal, active and engaging for learners.

ILYA HAS LEFT OPENAI !