ILYA HAS LEFT OPENAI !

Today we have AI systems that can write stories, systems that can create game maps, and systems that can play games. How about all of it at once? In this Word2World: Generating Stories and Worlds through Large Language Models researchers demonstrated a Word2World, a system that can generate complete stories and games - and play them!

Engineer leading scaling of ChatGPT etc. left OpenAI.

Google will have its new Trillium TPU processors manufactured on a TSMC 3nm or 4nm process, media report, adding Google’s design partner was Broadcom and the TPU is based on the Arm architecture.

Separately, the processor in Google’s Pixel 10 smartphones, due in 2025, will also be made using TSMC’s 3nm process.

Amazing application of VR combined with brain measurements in the interest of motor rehabilitation.

This is exactly the type of VR in healthcare learning application that I believe will disrupt healthcare.

This offers an extensive review of this literature.

From a neuroscience of learning perspective this is spot on.

VR broadly engages multiple motor learning centers in the brain in synchrony.

Combine that with signals that can be extracted in real-time from the brain and you have the potential for strong rehabilitation.

There is a ton more work to do but reviews like this can guide us.

AR/VR Market set for major growth?

IDC reports a promising turnaround for the AR/VR industry, with headset shipments expected to surge 44.2% to 9.7 million units in 2024.

Key Highlights:

1. Launches like the Meta Quest 3 & Apple's Vision Pro are not just capturing headlines but also pushing the boundaries for what's possible in AR and VR.

2. From virtual to mixed reality, companies are paving the way for authentic AR experiences, with a significant growth potential highlighted for both sectors through 2028.

Whats the future holding?

- Expected to reach 24.7 million units by 2028, expanding beyond gaming into business applications.

- Forecasted to grow at an 87.1% CAGR, reaching 10.9 million units by 2028.

Huggingface launched ZeroGPU, a shared infrastructure for indie and academic AI builders to run AI demos on Spaces

Technically speaking, ZeroGPU leverages Hugging Face's experience in hosting and serving more than 100 Petabytes monthly from the Hugging Face Hub.

ZeroGPU allows Spaces to run on multiple GPUs by making Spaces efficiently hold and release GPUs as needed (as opposed to a classical GPU Space that holds exactly one GPU at any time).

This architecture is also more energy-efficient since GPUs are shared rather than duplicated. ZeroGPU uses Nvidia A100 GPU devices under the hood.

Stages of Funding. Awesome Explainer by Spice Route Finance

1. This document provides an in-depth examination of the various stages of funding for startups, from initial rounds like pre-seed and seed funding to more advanced stages including Series A, B, C, and potentially Series D and E rounds.

2. Each stage is meticulously described, offering insights into the amounts typically raised, the sources of these funds, and the crucial aspects that attract investors at different phases of a startup's growth.

3. The document elucidates the journey of startup funding, beginning with personal investments and small angel rounds to substantial venture capital injections as companies scale and mature.

3. It discusses the dynamics and expectations at each funding stage, the challenges startups face, and the criteria investors consider before committing their capital.

4. The guide also highlights the potential pitfalls and successes that startups may encounter as they navigate through these critical stages of growth.

State of ZK Report (Q1-2024)

The "State of ZK Report for Q1 2024" presents a comprehensive overview of the significant developments, trends, and innovations within the Zero-Knowledge (ZK) ecosystem during the first quarter of 2024.

This report highlights the progress made in optimizing SNARK/STARK systems, the notable rollouts of new ZK technologies, and the ongoing integration of ZK solutions into various blockchain projects.

1. One of the key advancements this quarter was the rollout of EIP 4844 with Ethereum’s Mainnet Dencun upgrade, which significantly lowered Layer 2 (L2) gas fees.

2. This development, along with the introduction of new projects and products by existing ZK companies, underscores the dynamic nature of the ZK landscape.

3. Emerging segments such as ZK Prover Marketplaces, zkVMs (non-zkEVMs), and Fully Homomorphic Encryption (FHE) projects are beginning to take shape.

4. While FHE is not traditionally classified under the ZK umbrella, its synergy with ZK technology offers new possibilities for creating private and secure execution environments.

5. The report also details the growth of ZK prover marketplaces, including the announcement of Nebra and Succinct’s Prover Network, joining other established players like Nexus, =nil, RISC Zero, and Gevulot in decentralizing ZK proof generation.

6. Additionally, significant launches, such as the Succinct Processor 1 (SP1) zkVM, have spurred discussions on benchmarking standards for zkVM performance.

7. Furthermore, new Layer 1 (L1) blockchains continue to explore ZK integrations, with notable activities in projects like Celestia and Bitcoin, which have recently seen the emergence of ZK rollup teams.

8. The report concludes with an optimistic outlook for 2024, highlighting the continued momentum and innovation within the ZK space.

Google DeepMind just dropped a framework for how to evaluate future AI models as they are being created, avoiding potentially dangerous capabilities.

Apple has submitted a patent application as a “biosensing device” built into earbuds to measure “biological signal parameters from a user.”

Which biological signals, you ask?

1. Electroencephalography (EEG). In other words, the aim is to directly record the user’s brain waves from tiny sensors positioned within the ear canal.

2. Electromyography (EMG). This records muscle movements and the information can be used to help understand facial expressions and jaw movements related to emotion.

3. Electrooculography (EOG) tracks eye movements, particularly side-to-side.

4. Electrocardiogram (ECG) typically measures the electrical activity of the heart.

5. Galvanic skin response (GSR), which provides an indirect measure of emotional arousal – that is, the strength of an emotional response.

6. Blood volume pulse (BVP). This is measured using photoplethysmography and provides information about heart rate (HR) and heart rate variability (HRV).

You can read the application for yourself here.

AI in Finance by Deloitte

the most compelling use cases for Generative AI i across the finance organization:

-Financial Planning & Analysis
-Transactional Finance
-Controllership
-Strategic Finance
-Internal Audit
-Tax
-Treasury
- Investor Relations

For each of these domains, the paper explores how #Generative AI can address enterprise challenges in new ways, permit more and greater capabilities, and deliver advantages in efficiency, speed, scale, and capacity across the finance organization.

As with any type of AI, there are potential risks.

Deloitte’s Trustworthy AI framework is used to elucidate factors that contribute to trust and ethics in Generative AI deployments, as well as some of the steps that can promote #governance and #riskmitigation.

Trustworthy AI in this respect is: fair and impartial, robust and reliable, transparent and explainable, safe and secure, accountable and responsible, and respectful of privacy. To be sure, this collection of use cases is just a sample among myriad other applications, some of them yet to be conceived.

As Generative AI matures as a #technology and organizations move forward with using it for business benefit, we will likely see even more impressive and compelling use cases.

Here is a new player on the #surgical robot market. #Sony developed its own microsurgery assistance robot that is capable of automatic surgical instrument exchange and precision control.

"The system tracks the movements of a surgeon’s hands and fingers using a highly sensitive control device. The movements are then replicated on a small surgical instrument that operates similarly to the movement of the human wrist.

According to the company, the system addresses practical challenges in conventional surgical assistant robotics, such as interruptions and delays caused by manually exchanging surgical instruments."

AI Safety - 2024 AI Seoul Summit

AI Seoul Summit, jointed hosted by #UK and #SouthKorea. on 21 - 22 May 2024 represents an important opportunity to once again cement AI safety’s place on the international agenda.

The International AI Safety Report has been released ahead of the AI Seoul Summit which aims to give policymakers across the globe a single source of information to inform their approaches to AI safety.

The AI Safety Report recognises that advanced AI can be used to boost wellbeing, prosperity, and new #scientific breakthroughs – many of which have already been seen in fields including healthcare, drug discovery, and in how we can tackle climate change.

But it notes that like all powerful technologies, current and future developments could result in harm. For example, malicious actors can use AI to spark large-scale disinformation campaigns, fraud, and scams. Future advances in advanced AI could also pose wider #risks including labour market disruption, and economic power imbalances and inequalities.

However, the report highlights a lack of universal agreement among AI experts on a range of topics, including both the state of current AI capabilities and how these could evolve over time. It also explores the differing opinions on the likelihood of extreme risks which could impact society such as large-scale unemployment, AI-enabled terrorism, and a loss of control over the #technology.

Attendees will be able to use the interim International AI Safety Report to further the discussions which were kickstarted at November’s AI Safety Summit in the UK. A final edition of the report is expected to be released ahead of next round of discussions on AI safety which will be hosted by France.

Virtual 3D spaces are playing larger roles in the lives of consumers and employees as evidenced by growing operations among the metaverse's early movers such as Roblox and Siemens.

And the potential of this market is drawing other entrenched media and tech players, such as Apple, Walt Disney and Microsoft.

In other words, the flywheel is in motion, and the metaverse is primed to transition from a solution in search of a problem to a substantive piece of the media and productivity landscape over the five-year forecast, according to the latest forecast from S&P Global Market Intelligence.

Microsoft CEO Satya Nadella said AI Copilot+ PCs beat Macs on Monday as the cloud giant unveiled new tablets and laptops built for Copilot, while Dell also unveiled its first wave of Copilot+ AI laptops optimized for AI, noting the laptops use Qualcomm processors.

A wave of AI PC announcements are expected during the Microsoft Build conference this week and Computex Taipei 2024 in early June.

Copilot+ PCs can run AI workloads up to 20x faster and 100x more efficiently than traditional PCs, and are 58% faster than a Macbook Air M3.

One of the biggest AI features launched in the Copilot Plus PCs is called 'Recall'. This is a Black Mirror episode.

It allows users to search and recall anything they’ve seen and interacted with on their computer screens, with natural language.

Copilot is also getting a new screen-sharing feature powered by GPT-4o

It allows AI to watch, hear, and understand what a user is doing on their computer and answer questions in real time.

#ASML and #TSMC have ways to disable the world's most advanced #chipmaking machines in the event #China invades #Taiwan

Officials from the US government have privately expressed concerns to both their Dutch and Taiwanese counterparts about what happens if Chinese aggression escalates into an attack on the island responsible for producing the vast majority of the world’s advanced semiconductors. 

ASML reassured officials about its ability to remotely disable the machines when the Dutch government met with the company on the threat. The Netherlands has run simulations on a possible invasion in order to better assess the risks.

The remote shut-off applies to ASML’s line of extreme ultraviolet machines, known as EUVs, for which TSMC is its single biggest client. #EUVs harness high-frequency light waves to print the smallest microchip transistors in existence — creating chips that have artificial-intelligence uses as well as more sensitive military applications. China has never been able to buy EUV machines from ASML.

Two more great papers on AI agents:

1. Researchers argues that we need robust identifiers for agents.

2. Noam Kolt provides the most interesting analysis, principal-agent analysis and agency law.

Images that Sound: Composing Images and Sounds on a Single Canvas

This paper introduces an inference-time procedure that generates images that are also spectrograms corresponding to the prompt. It uses a latent image and audio diffusion model with same latent space (Stable Diffusion v1.5 and Auffusion) and denoise the same latent with both.

Code.

Anthropic announced that they’ve gotten dictionary learning working on Sonnet, extracting millions of features from one of the best models in the world.

This is the first time this has been successfully done on a frontier model.

For context, the goal of dictionary learning is to untangle the activations inside the neurons of an LLM into a small set of interpretable features.

Sanofi and OpenAI announced a partnership focused on customizing large language models for drug development

Another (early) example of expanding relationships between AI & biopharma worlds.