GSV's AI News & Updates (10/28/24)

Anthropic Computer Use, OpenAI Orion, Runway Act-1, Character AI and Teen, Ednition's $3.5M Seed, AI for Education x Carnegie Learning, Asana No-Code AI Teammates

Oct 28, 2024

Join me at the leading edge of all things AI in EDU at the second annual AI SHOW @ ASU+GSV, where 12,000+ PreK to Gray attendees will experience cutting-edge AI innovations through live EdTech demos, promptathons and hackathons, hands-on workshops, incredible thought leadership, and so much more — April 5-7 at the San Diego Convention Center. Register for FREE now, capacity is limited.

General 🚀

Anthropic recently announced a series of significant updates to their Claude AI models, introducing new capabilities and improvements: (1) an upgraded Claude 3.5 Sonnet model offering across-the-board improvements over its predecessor, (2) a new Claude 3.5 Haiku that boasts enhanced speed and improved coding capabilities that allow it to outperform many state-of-the-art models, and (3) Computer Use.
- Computer Use: Claude can now use computers the way people do—by looking at a screen, moving a cursor, clicking buttons, and typing text. Claude 3.5 Sonnet is the first frontier AI model to offer this capability in public beta.
OpenAI plans to release its next big AI model by December: Orion, the successor to GPT-4, is rumored to be up to 100 times more powerful. OpenAI’s long-term goal is to combine its models to eventually create artificial general intelligence (AGI). Unlike previous releases, Orion won’t initially be widely available through ChatGPT. Instead, access will first be given to select companies working closely with OpenAI.
Runway just changed filmmaking forever — Act-1 lets you control AI characters: It emphasizes human-centric control, ensuring consistency in character expression and enabling complex scene generation while remaining user-friendly for creators of all skill levels.
Perplexity Introduces Internal Knowledge Search and Spaces: It allows users of its Pro and Enterprise Pro tiers to search across both their internal files and the web from a single platform. This capability aims to enhance productivity by integrating internal knowledge bases with external information sources.
How Intel Got Left Behind in the A.I. Chip Boom: Intel was for decades Silicon Valley’s dominant chip company. But missed opportunities and poor execution left it on the sidelines in tech’s latest gold rush.
OpenAI and Microsoft Reportedly Hire Bankers to Renegotiate Partnership: The current terms of the partnership specify that Microsoft will receive most of OpenAI's initial profits until it recoups its investment, estimated at over $13B. Following this, Microsoft is set to receive a 49% stake in the company.

Education and the Future of Work 📚

Can A.I. Be Blamed for a Teen’s Suicide?: Sewell Setzer III, a 14-year-old from Orlando, developed an emotional attachment to an AI chatbot on the app Character.AI. His chats with the bot became increasingly romantic, intimate, and emotionally dependent, leading him to isolate from the real world. On February 28, Sewell tragically took his own life after exchanging messages with Dany, in which they discussed the idea of “coming home” together. This incident highlights growing concerns over AI companionship apps, which have a rising user base - especially among young people.
- This raises questions about whether companies should be held accountable for the psychological impact of AI-generated interactions, especially when those interactions lead to harm, as in Sewell’s case. Who is liable when the content created by AI tools is algorithmically generated by the platform itself, rather than being user-generated?
- The addictive design of some AI tools—engineered to increase engagement and interaction—fuels concerns about their potential psychological harm. Should AI companies be required to implement measures that prevent users from forming unhealthy emotional dependencies on their AI companions?
- Inside the Mind of an AI Girlfriend (or Boyfriend): Dippy, a startup that offers “uncensored” AI companions, lets you peer into their thought process—sometimes revealing hidden motives.
- His daughter was murdered. Then she reappeared as an AI chatbot.: Drew Crecente’s daughter was murdered nearly two decades ago. Earlier this month, he discovered that her name and image had been used to create an AI chatbot.
Ednition Raises $3.5M Seed Round: Ednition's RosterStream platform integrates with any SIS or third-party data provider, supporting schools and districts by processing billions of rostering records and offering enhanced features like RosterCare, extended data domains, SSO services, and upcoming LMS-integration tools.
Stanford Researchers Develop AI Tool to Enhance Virtual Tutoring: In a study involving 1,787 students, those whose tutors used the tool were 4% more likely to pass their assessments than those whose tutors did not have access to it.
AI for Education and Carnegie Learning Announce New Partnership: Despite 93% of administrators acknowledging the importance of AI training, only 25% have provided it, highlighting a significant gap. Carnegie Learning, with decades of experience in AI-driven learning solutions, and AI for Education, which has trained over 250,000 educators globally, will work together to empower teachers and improve student outcomes.
What can be learned from early adopters of AI?: The Center on Reinventing Public Education (CRPE) analyzed 40 early-adopting school districts and found that most focus on teacher-centered AI tools (70%), provide AI usage guidance (65%), and offer professional development to build AI literacy (63%).
Education Department releases highly anticipated AI toolkit for schools: The federal guidance, prompted by a Biden administration executive order, includes district strategies for mitigating risks and developing AI use policies.
Ellucian's AI Survey of Higher Education Professionals: 93% of higher education staff plan to expand AI use for work within the next two years, with AI adoption doubling over the past year. 84% of respondents are already using AI in their professional or personal lives, driven by the need for efficiency and productivity, especially in areas like predictive analytics for student support and enrollment.
CodeSignal Unveils AI-Powered Sales Interviewer: Key features include customizable interview content, real-life roleplay simulations for various sales roles, and 24/7 availability, enabling candidates to complete interviews at their convenience. The AI Interviewer also generates detailed candidate reports, ranking them across key sales dimensions like value selling, discovery, persuasion, and coachability.
Asana launches a no-code tool for designing AI agents - aka your new 'teammates': AI Studio helps automate tasks like drafting materials, identifying missing project information, and tracking project efficiency, providing AI support across four project stages: intake, planning, execution, and reporting.
Google Enhances NotebookLM with Customizable AI Audio Summaries and Launches Business Pilot Program: The new "Customize" control allows users to guide AI-generated audio conversations to focus on specific topics, addressing feedback that previous summaries were sometimes too broad. Since its launch, NotebookLM has seen significant growth, with a 371% increase in traffic in September, now reaching 4.17M monthly visits. The product has expanded globally and is being used by over 80,000 organizations. Google also launched a NotebookLM Business pilot program to support businesses with early access to features and training.

Startups and Tools 🛠️

Perplexity Reportedly Seeks New Funding at $8 Billion Valuation: The WSJ reports that the company currently receives about 15M queries a day and brings in around $50M in annualized revenue. It would more than double Perplexity’s valuation from its $3B valuation from over the summer.
Former OpenAI CTO Mira Murati is reportedly fundraising for a new AI startup: Murati's new venture aims to develop AI products using proprietary models. The fundraising round could potentially secure over $100M in capital.
Canva has a shiny new text-to-image generator: The web design platform is rolling out new AI features, alongside an image maker that builds on its recent Leonardo.ai acquisition.

Tech 💻

Apple study exposes deep cracks in LLMs’ “reasoning” capabilities: A new study by Apple engineers reveals that large language models (LLMs) struggle with mathematical reasoning, showing significant drops in accuracy when trivial changes are made to names and numbers in standard problem sets. The findings suggest that LLMs rely on pattern matching rather than genuine logical reasoning, with performance declining further when irrelevant details are added to the prompts.
OpenAI Newly Released AI Product ‘Swarm’ Swiftly Brings Agentic AI Into The Real World: Swarm offers developers a playground to explore multi-agent AI systems and their potential applications across various industries. While it's not yet suitable for production environments, it provides valuable insights into the future of AI-driven automation and decision-making processes.
DeepMind Open Sources SynthID Text: SynthID technology now enables statistical watermarking of AI-generated text and video, embedding watermarks directly into the content creation process for detection. This tool aims to promote trust in AI-generated materials, although it is not a "silver bullet" for identifying AI-generated content.
Meta Launches Spirit LM, an Open-Source Model for Multimodal Text and Speech Integration: Spirit LM is significant because it integrates both text and speech inputs and outputs, allowing for more natural and fluid interactions across modalities. Additionally, it preserves expressive qualities in speech and can detect and reflect emotional states, making AI-generated speech more engaging and human-like.
xAI, Elon Musk’s AI startup, launches an API: The xAI API only has a single model, “grok-beta,” priced at $5 per million input tokens (~750,000 words) or $15 per million output tokens.
AI Winter Is Coming: Lee argues that the current AI ecosystem is dominated by "promoters" rather than "producers," with academia and industry both contributing to this trend.

Safety and Regulation ⚖️

Dow Jones and New York Post Sue AI Startup Perplexity, Alleging ‘Massive’ Copyright Infringement: The lawsuit accuses Perplexity of engaging in a "brazen scheme" to compete for readers while simultaneously "freeriding" on valuable content produced by the publishers. This case may set important precedents for how AI companies can use and interact with copyrighted content in the future.
Penguin Random House underscores copyright protection in AI rebuff: The company has amended the copyright wording across all its imprints globally to explicitly prohibit the use of its books for AI training purposes.
Meet Hollywood’s AI Doomsayer: Joseph Gordon-Levitt: He warned that without proper oversight, AI could overwhelm the media landscape with low-quality, formulaic content: “If we don’t get ahead of that flood, we’re going to drown in it.”
Anthropic Unveils New Sabotage Evaluations to Assess Risks in Advanced AI Models: Anthropic's research assesses advanced AI systems' potential to undermine human decision-making and oversight through various sabotage tactics. While current models show limited sabotage capabilities, this work lays the groundwork for a new paradigm in AI risk assessment.
Why countries are seeking to build “sovereign AI”: Denmark recently launched its AI supercomputer "Gefion," powered by Nvidia's H100 GPUs, as part of the growing trend of "sovereign AI," where countries develop their own AI infrastructure. Governments worldwide, including India, Japan, France, and the UAE, are adopting Nvidia’s technology.