AI trends to watch into 2025 From Edge AI to AI chips, these were the key AI innovations at Embedded World 2024.

Japanese talent pushed to AI chip firm Tenstorrent by government The Japanese government will send engineers to be trained in RISC-V and AI at Tenstorrent. The latter gets revenue and talent. Japan hopes to revitalise its semiconductor industry. RISC-V technology gets a boost.

UAE: Autonomous AI for Oil Production ADNOC, in partnership with G42, Microsoft, and AIQ, will deploy autonomous agentic AI to revolutionise the energy sector. UAE officials see this AI, which autonomously optimises operations and forecasts with high accuracy, as central to diversifying the economy beyond oil.

New plan needed for AI nuclear power US tech giants looking to quickly power massive artificial intelligence centers will need to come up with new strategies after US regulators quashed Amazon.com Inc.’s effort to tap a Pennsylvania nuclear facility.

Trump's win signals impact to everything from AI safety to GPU costs The future of AI regulation may now hinge on individual states, whiletariffs on GPUs certainly won't do researchers any favours.

Interpretability vs. Explainability in AI What they are, their differences, and why they matter.

Merging Deep Learning Accuracy with Model Interpretability SEE-Net offers a breakthrough in explainable AI by integrating a Deep Neural Network with a simplified neural network for interpretable predictions. This architecture enables SEE-Net to retain DNN-level accuracy while offering transparency akin to statistical models, using co-supervision to create an "almost white-box" model. Tests on image and tabular data reveal SEE-Net’s potential to maintain both prediction power and user interpretability across applications.

UK Government: Assuring a Responsible Future for AI ‘Assuring a Responsible Future for AI’ report outlines three key areas for building a robust AI assurance landscape in the UK: 1. Creating a ‘one-stop shop’ platform to drive demand for AI assurance services 2. Improving AI assurance quality with industry & the AI Safety Institute 3. Developing a responsible AI terminology tool with NPL and US NIST

Accuracy, Bias in AI Concerns Most CEOs Nearly half of CEOs are concerned about accuracy and bias when it comes to AI, according to a new survey by the IBM Institute for Business Value on AI governance.

Enhancing AI Reasoning by Addressing Softmax Limitations in Sharp Decision-Making An adjustable temperature mechanism inside the softmax function as a solution to find the most pertinent data points as the problem size increases.

New Hybrid Algorithm Advances Well Pattern Optimisation in Field Development This study introduces an innovative hybrid algorithm for optimising well locations and settings in oil fields, combining genetic and particle swarm algorithms. By allowing suboptimal solutions to improve through adaptive velocities, the new method surpasses traditional algorithms, offering faster, cost-effective simulations and better outcomes for complex field scenarios. The results showcase enhanced optimisation potential, particularly in high-dimensional planning processes for well placement and resource maximisation.

Optimising Compact IFA Antennas with Pixelated Design for High-Performance Wireless Applications This study presents a novel approach to compact antenna design using pixelated, Einstein Hat-shaped elements in Inverted-F Antennas (IFAs), optimised by genetic algorithms. Compared to traditional IFA designs, the optimised antenna demonstrates improved gain, efficiency, and reduced reflection, proving effective for spatially constrained systems in IoT and other advanced wireless applications. Measurements in anechoic chambers confirm the enhanced performance, marking this as a promising technique for next-gen compact antennas

Android-Lab GitHub: Benchmarking autonomous AI agents on Android OS.

Altman-backed Oklo wants to make clean nuclear energy more accessible Energy and power grid constraints look to be the biggest hurdles for Big Tech to overcome in the industry's wider buildout of AI data center infrastructure. Oklo Inc. (OKLO) is one of these names benefitting from the trend, its stock having jumped nearly 200% over the past month. The nuclear startup is backed by OpenAI CEO Sam Altman, who is also Oklo's chairman.

Meet Literal Labs @ Web Summit, today British edge AI startup, Literal Labs, is exhibiting at Libson's Web Summit today. Meet them at A235 in the startups' space.

AI World Congress 2024, London AI World Congress comes to Kensington this November with keynote speakers from AWS, Oracle, WPP, and SingTel.

AI Leaders Forum 3 December, London.

Apple exploring smart glasses market (again) After pursuing self-driving cars and Vision Pro, smart glasses are back on the cards for Apple newly motivated by the success of Ray-Bans / Meta.

Saudi vs UAE: $100 Billion AI hub with Google An AI hub being built by Saudi Arabia and Google serves as the starting point for a broader initiative. The companies are planning to invest between $5 billion and $10 billion initially.

AI and US politics Under Trump: AI Executive Order to be repealed, CHIPS faces challenges. DOD contractors, such as Palantir, undoubted winners.

Writing as a way of thinking Generative AI and its impact on our ability to reason.

AI patent to counter hallucinations Microsoft patent for a ‘response-augmenting system’ engineered to combat AI hallucinations by double-checking responses against real-world information.

This week's issue presented by Literal Labs.

Was this email forwarded to you? Sign up for free here.