Twice a week, Data Points brings you the latest AI news, tools, models, and research in brief. In today’s edition, you’ll find:
- A new safety investigation of Tesla’s self-driving tech
- Perplexity adds collaborative spaces and internal search
- Nvidia’s new Nemotron model scores big on benchmarks
- Telegram’s subculture of explicit deepfake generators
But first:
DeepSeek unveils Janus, a versatile AI model for text and images
DeepSeek released Janus, a new AI framework that handles both understanding and generating multimodal content. The system uses separate visual pathways but a single transformer architecture, improving on previous approaches and increasing flexibility. Janus outperforms earlier unified models and is competitive with specialized models at multiple vision understanding and generation benchmarks. The model is available for download on a permissive open license. (GitHub)
SambaNova and Gradio partner to simplify AI model deployment
SambaNova Systems and Gradio announced an integration that allows developers to access high-performance AI models with minimal code. The partnership enables users to create web applications powered by AI models running on SambaNova’s hardware using Gradio’s gr.load() function, simplifying the process of building chat and other interfaces. This collaboration makes it easier for developers to work with high-speed inference infrastructure in AI web applications. (GitHub)
Safety regulator probes Tesla’s self-driving tech after collisions
The U.S. National Highway Traffic Safety Administration launched an investigation into Tesla’s supervised full self-driving technology, examining four collisions, including one fatal pedestrian accident. The probe focuses on whether the software has adequate safeguards to ensure drivers can retake control when necessary. This could pose a significant challenge to Tesla’s ambitious plans for autonomous vehicles that rely on cameras rather than a combination of sensors. (NHTSA and The New York Times)
Perplexity adds internal search and collaborative tools for pro users
Perplexity introduced Internal Knowledge Search for Pro and Enterprise Pro users, allowing them to search across both web content and internal files simultaneously. The company also launched Perplexity Spaces, AI-powered collaboration hubs that teams can customize for specific research and organizational needs. These new features enable Perplexity’s industry users to search internal data alongside public information, enhancing due diligence, sales processes, and employee support. (Perplexity)
Nvidia’s new AI model gets top marks on arena/chat benchmarks
Nvidia created a new AI called Llama-3.1-Nemotron-70B-Instruct that scored higher than GPT-4 and Claude 3.5 on three important tests. These tests measure how well AI understands and follows instructions, with Nvidia’s model scoring 85.0 on Arena Hard, 57.6 on AlpacaEval 2 LC, and 8.98 on GPT-4-Turbo MT-Bench. Nvidia used special training methods to teach their AI, including reinforcement learning from human feedback and use of its own Nemotron reward model. (Nvidia)
Explicit AI deepfake bots flourish on Telegram
Wired identified at least 50 Telegram bots claiming to generate explicit photos or videos of individuals with minimal user input. These bots collectively report over 4 million monthly users, with some individual bots boasting hundreds of thousands of users each. The persistence and prevalence of these tools on Telegram highlights the platform’s role as a major hub for deepfake creation, even though the identified bots likely represent only a fraction of the total number available. (Wired)
Still want to know more about what matters in AI right now?
Read last week’s issue of The Batch for in-depth analysis of news and research.
Last week, Andrew Ng argued that we should consider geoengineering to be an important potential tool to mitigate climate change.
“All these downsides should be balanced against the reality that people are dying. I’m moved by meteorologist John Morales’ emotional account of the havoc caused by Hurricane Milton. The New York Times quoted him as saying, ‘It claims lives. It also wrecks lives.”
Read Andrew’s full letter here.
Other top AI news and research stories we covered in depth: Malaysia helps drive a data center boom driven by its strategic location, natural resources, and investor-friendly policies; the U.S. launches Operation AI Comply to crack down on AI applications that overpromise and underdeliver; a new report highlights the contending forces shaping AI, including the battle between open and proprietary technology; and researchers introduce a better text embedding model with adapters specialized for tasks like retrieval, clustering, and text classification.