🥹 Anthropic Hires "AI Welfare" Researcher

Anthropic's pioneering move to hire a dedicated AI welfare researcher underscores growing ethical considerations in AI development.

Author

Jack Lajoie
November 12, 2024 • read time ~ 9 minutes

The Daily Current ⚡️

Welcome to the creatives, builders, pioneers, and thought leaders ever driving further into the liminal space.

Anthropic has taken a pioneering step by hiring its first full-time AI welfare researcher, highlighting the growing ethical considerations in AI development. Simultaneously, the AI adoption landscape is evidently experiencing a slowdown, as observed in a Slack-commissioned study. Meanwhile, regional search engines like Ecosia and Qwant are set to actually challenge the dominance of tech giants like Google with new initiatives. Additionally, Google DeepMind's release of the AlphaFold3 code has sparked discussions on open science, and Baidu is expanding its AI offerings with advanced tools, intensifying the competitive race in the AI market.

 🔌 Plug Into These Headlines:

Anthropic has hired Kyle Fish as its first full-time AI welfare researcher, exploring the potential moral status of future AI systems. This move reflects growing concerns in the AI industry about the ethical implications of creating potentially conscious or morally relevant AI, and the need to consider their welfare. Fish’s role involves investigating complex philosophical and technical questions surrounding AI welfare, including what capabilities might make an AI system worthy of moral consideration and how companies should respond to these possibilities.

  • A recent report co-authored by Fish argues that AI welfare is no longer just a sci-fi concept.

  • There are concerns about potentially creating and mistreating vast numbers of suffering AI beings.

  • The field of AI welfare research is still in its early stages, with many unanswered philosophical and practical questions.

  • Fish is focusing on fundamental questions about AI welfare and potential practical steps.

🤖 As AI systems become more advanced, the line between machine and sentient being blurs, forcing us to grapple with unprecedented ethical dilemmas in technology development.

The latest Workforce Index from Slack indicates that the initial AI hype is subsiding among global desk workers. While executives remain bullish on AI, with 99% planning investments, worker adoption has plateaued, and excitement has decreased. The survey reveals underlying issues such as uncertainty about AI norms, fears of being perceived negatively for using AI, and a significant lack of training, all contributing to the cooling sentiment towards AI in the workplace.

  • As of August 2024, 99% of executives plan to invest in AI in the coming year, showing full commitment to AI.

  • Between March and August 2024, AI adoption rates stalled in some countries, with France seeing only 2% growth and the U.S. just 1%.

  • Excitement around AI dropped 6 percentage points globally (from 47% to 41%) among desk workers over a three-month period.

  • Nearly half (48%) of all desk workers would be uncomfortable admitting to their manager that they used AI for common workplace tasks.

  • 61% of desk workers have spent less than five hours total learning how to use AI, indicating a persistent lack of training.

🎢 The AI hype cycle in the corporate world is entering a reality check phase, as the initial excitement gives way to practical challenges of implementation and acceptance. After employees led the initial charge, often adopting AI without executive approval, it seems now that executives are finally onboard the employees have already grown tired of the trend.

Ecosia and Qwant, two European search engines, are joining forces to create their own web index called European Search Perspective (EUSP). This collaboration aims to reduce their reliance on Google and Microsoft’s Bing for search results, allowing them to tailor results according to their company missions and European preferences. The joint venture, equally owned by both companies and based in Paris, plans to begin testing revamped search results in France in early 2025 and in Germany by the end of that year.

  • The initiative is driven by recent technological advances and new EU regulations limiting the power of tech giants.

  • The companies are open to outside funding and licensing their index to other companies, including for AI training.

  • EUSP will be equally owned by Ecosia and Qwant and based in Paris.

  • Testing of new search results is planned for France in early 2025 and Germany by the end of 2025.

  • Ecosia has about 1% market share in France and Germany, with 20 million users globally, while Qwant reports 6 million users.

🌍 This collaboration highlights a growing potential for regional search engines to challenge tech giants as they face increased regulations. Perhaps in markets where “local knowledge” and “cultural sensitivity” can provide a competitive edge, they might have a chance.

Google DeepMind has released the complete code for AlphaFold3, its advanced protein prediction software, following a six-month delay that sparked criticism from the scientific community. This release allows researchers to access the computational model and its weights, enabling them to evaluate and build upon the technology more effectively. While many scientists are pleased with this development, some express frustration over the initial withholding of the code, which contradicted established open science practices.

  • Google DeepMind released the full code for AlphaFold3 on November 11, 2024, six months after its initial announcement in May 2024.

  • The release includes the computational model on GitHub with a noncommercial license and “weights” available to academics through an application.

  • The initial withholding of the code faced criticism from the scientific community, leading to an open letter with hundreds of signatures.

  • Nature journal waived its requirement for sharing full code, citing potential biosecurity implications.

  • Some researchers express frustration over the 6-month delay, calling it “unacceptable” despite welcoming the release.

🔍 The delayed release of AlphaFold3’s code exposes the challenges of maintaining scientific integrity in a competitive landscape where commercial interests often clash with academic norms.

Baidu has expanded its AI portfolio with two new tools: ERNIE-VilG 2.0, an advanced text-to-image generator, and a no-code app builder. These additions to Baidu’s AI Cloud platform aim to enhance the company’s competitive edge in the global AI market, offering users improved image generation capabilities and simplified AI application development.

  • Baidu expands its AI lineup with two new tools: a text-to-image generator and a no-code app builder.

  • The text-to-image generator, ERNIE-VilG 2.0, is an upgrade to Baidu’s existing image generation technology.

  • These new offerings are part of Baidu’s strategy to strengthen its position in the rapidly evolving AI landscape.

  • The tools are integrated into Baidu’s AI Cloud, enhancing the company’s suite of AI services.

🌐 Baidu’s latest AI tools underscore the intensifying race among tech giants to dominate the AI market, with user-friendly interfaces becoming a key battleground.