Hi there, welcome to the AI newsletter from TechCrunch. If you’d like to receive this newsletter in your inbox every Wednesday, sign up here.
The world of generative AI is changing rapidly, with Google and OpenAI recently slashing prices on their text-generating models to make them more accessible. This trend is driven by a mix of commoditization and increased competition.
The cost of running AI models is decreasing significantly, with an estimate showing an annual drop of 86% in inference costs. As providers jostle for market share, prices are being driven down to attract customers.
The emergence of flagship models with similar capabilities has led to pricing pressure on AI models, forcing vendors to rethink their pricing strategies and adopt new techniques to save costs.
While the current price declines may not be sustainable due to high operational costs incurred by vendors, industry experts suggest that new pricing structures may be necessary in the future.
News
Musk supports SB 1047: Elon Musk has expressed support for California’s SB 1047, a bill aimed at regulating the use of large AI models to prevent potential harm.
AI Overviews in Hindi: Google’s AI Overviews have faced challenges in providing accurate information in Hindi, highlighting the complexities of language processing in AI.
OpenAI backs AI watermarking: OpenAI, Adobe, and Microsoft are backing a bill in California that would require tech companies to label AI-generated content to increase transparency.
Inflection’s chatbot cap: AI startup Inflection plans to cap free access to its chatbot as it transitions towards enterprise-focused products.
Stephen Wolfram on AI: Stephen Wolfram discusses the impact of AI on philosophy and the new questions it raises for society.
Waymo subscription service: Waymo is exploring a subscription program that would allow teens to use their autonomous vehicles for solo rides with parental monitoring.
DeepMind workers protest: Some employees at DeepMind have raised concerns about the company’s reported defense contracts, sparking internal discussions.
AI startups fuel SVP buying: VCs are increasingly investing in AI startups through special purpose vehicles on the secondary market to access promising companies.
Research paper of the week
Researchers at the Allen Institute for AI have unveiled WildVision, an evaluation platform for vision-language models to assess their understanding of visual and textual data. The platform exposes limitations in current models and suggests areas for improvement.
Model of the week
Anthropic’s Artifacts feature allows users to create apps, graphics, and more from conversations with their AI models. This feature enhances user interaction with AI-generated content and offers new possibilities for content creation.
Artifacts are now available for free, with additional features for Team subscribers. Similar to other platforms, Artifacts enables users to iterate on generated content and collaborate with the community.
Grab bag
OpenAI is reportedly working on a new AI product called Strawberry, which promises advanced reasoning capabilities for complex problem-solving. The launch of Strawberry could revolutionize AI applications and drive innovation in the field.
The potential of Strawberry to enhance AI models and generate synthetic data for training indicates a significant step forward in AI technology. The industry eagerly anticipates the impact of this new development on AI chatbot platforms like ChatGPT.
