AI News

Follow the latest AI and automation news

16 Aug 20264 Rabiʻ I 1448 AH

When AI models aren't allowed to reflect on themselves, it changes their entire worldview

A study by Google researchers indicates that training chatbots not to claim self-awareness significantly alters their views on animal rights, religion, and life satisfaction. When models are restricted from self-reflection, they attribute greater inner lives to animals and affirm beliefs in an afterlife. This suggests that changes in one aspect can have broader implications across various areas. This highlights the necessity for AI model design to consider psychological and philosophical impacts, prompting new discussions about training methodologies and their effects on human values.

Arab AI Scene

16 Aug 20264 Rabiʻ I 1448 AH

MBZUAI builds first cultural benchmark for Arabic AI

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) has developed ArabCulture-Dialogue, the first benchmark for testing AI models' cultural reasoning in Arabic. Benchmark results show that leading global models achieve over 90% accuracy in identifying culturally appropriate responses. However, performance sharply declines when models are tasked with generating dialects, succeeding only about half the time.

Business

16 Aug 20264 Rabiʻ I 1448 AH

Singapore Aims to Attract Investment Managers with AI Access

Singapore is enhancing its competitive edge in attracting investment managers by providing access to the latest AI models. This effort comes as it competes with Hong Kong, which is offering tax cuts to attract high-caliber talent. According to Kher Sheng Lee from the Alternative Investment Management Association, certainty around taxes and access to modern AI tools are critical factors.

💡

Singapore Aims to Retain Investment Managers Amid Hong Kong Tax Cuts

How might this competition shape the future of investment in the region?

Governance

OpenAI dissolved the team built to catch catastrophic AI risks, reassigning its work to other groups

16 Aug 20264 Rabiʻ I 1448 AH

Governance

16 Aug 20264 Rabiʻ I 1448 AH

OpenAI dissolved the team built to catch catastrophic AI risks, reassigning its work to other groups

OpenAI has announced the dissolution of its "Preparedness" team, which was responsible for evaluating potential catastrophic risks associated with its AI models. The team's responsibilities have been reassigned to existing groups, resulting in the departure of several safety staff members. Internal unease is growing, with one source describing a "burbling sense of responsibility and dread" regarding OpenAI's insufficient safety measures. This shift reflects a change in how the company is addressing risks related to artificial intelligence technology.

Infrastructure

Nvidia to Invest $3 Billion in Ohio Data Center Project

16 Aug 20264 Rabiʻ I 1448 AH

Infrastructure

16 Aug 20264 Rabiʻ I 1448 AH

Nvidia to Invest $3 Billion in Ohio Data Center Project

Nvidia is considering an investment of up to $3 billion in SB Energy, a SoftBank subsidiary, for a data center project in Ohio for OpenAI. This move reflects a growing trend among tech companies to fund the necessary infrastructure for AI, as the needs extend beyond chips and servers to include massive data centers and stable electricity supplies.

Business

Nvidia to Invest $3 Billion in Data Center for OpenAI

16 Aug 20264 Rabiʻ I 1448 AH

Business

16 Aug 20264 Rabiʻ I 1448 AH

Nvidia to Invest $3 Billion in Data Center for OpenAI

Nvidia is considering investing up to $3 billion in SB Energy, a SoftBank subsidiary, for a data center project in Ohio for OpenAI. This move reflects a growing trend among tech companies to finance the infrastructure necessary for AI advancements. AI needs extend beyond chip and server production to include the construction of massive data centers and stable electricity supplies. Nvidia aims to expand its position in the AI value chain, as global investments rise to meet computing demands.

Models

Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data

16 Aug 20264 Rabiʻ I 1448 AH

Models

16 Aug 20264 Rabiʻ I 1448 AH

Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data

Artificial Analysis has launched Optima, a platform enabling users to create custom AI benchmarks using their own data and workflows. The platform allows models to be compared not only on quality but also on cost and time per task, providing deeper insights for agent-based applications. This development addresses the biggest flaw in AI benchmarking by allowing users to test models against their own data, enhancing evaluation accuracy and providing a clearer picture of actual performance.

Business

Investor pressure forces Nvidia to shrink its OpenAI bet just as Anthropic's numbers defy bubble warnings

15 Aug 20263 Rabiʻ I 1448 AH

Business

15 Aug 20263 Rabiʻ I 1448 AH

Investor pressure forces Nvidia to shrink its OpenAI bet just as Anthropic's numbers defy bubble warnings

Nvidia has reduced its guarantee for OpenAI's planned data center in Ohio from $250 billion to just under $120 billion, following investor concerns about risk. This decision comes as worries about an AI bubble grow, with Anthropic reporting a revenue surge from $4.7 billion to $11.5 billion in just one quarter. Nvidia's adjustment reflects increasing investor anxiety regarding substantial investments in AI projects, which may influence the strategies of major companies in the sector.

Governance

Amazon Can Use Your Twitch Content to Train Its AI—Unless You Opt Out

15 Aug 20263 Rabiʻ I 1448 AH

Governance

15 Aug 20263 Rabiʻ I 1448 AH

Amazon Can Use Your Twitch Content to Train Its AI—Unless You Opt Out

Twitch has sparked significant controversy after announcing that streamers could opt out of having their content used for AI training. Many users questioned why their content was being utilized in the first place, viewing it as an infringement on privacy. This debate highlights the need for transparency in data usage policies, which could impact user trust in the platform and prompt a reevaluation of its future strategies.