Google's Made by Google 2024 event has concluded, featuring announcements and reveals on AI, Pixel 9 lineup, Gemini AI, Nest, Pixel Watch, and Pixel Buds. The event started at 10 a.m. PT and is available to watch on livestream archive. Google's focus is on AI, with Gemini AI being incorporated into devices.
AIGoogle | Rating: 53 | 2024-08-20 09:15:37 AM |
EliseAI, a company founded by Minna Song and Tony Stoyanov, has raised $75 million in a Series D round, valuing the startup at $1 billion. The company develops AI-powered property management tools for landlords, aiming to automate communication with current and prospective tenants. Song, the CEO, recognized inefficiencies in the rental and leasing industry while working as an administrative assistant at a residential real estate firm in New York City.
AIEliseAI | Rating: 53 | 2024-08-20 09:15:36 AM |
MIT researchers have developed an AI risks repository to guide policymakers and stakeholders in crafting laws to regulate AI. The repository aims to identify specific risks associated with AI systems, including those with control over critical infrastructure, scoring exams, sorting resumes, and verifying travel documents. The EU AI Act and California's SB 1047 are examples of laws that struggle to define which risks to cover.
AIMIT | Rating: 53 | 2024-08-20 09:15:36 AM |
Elon Musk-owned xAI's Grok AI model can now generate images on X social network, with access limited to Premium and Premium+ users. The model uses FLUX.1 by Black Forest Labs to create images. Early images suggest that Grok's image generation feature does not have guardrails around creating images of political figures, but the company may put limits in place to avoid regulatory scrutiny.
AIxai | Duplicated with: 1 | Rating: 47 | 2024-08-20 09:15:36 AM |
Researchers from the University of Bath and University of Darmstadt conducted thousands of experiments to test the ability of several generative AI models, including those in Meta's Llama family, to complete tasks they hadn't come across before. The study found that the models could superficially follow instructions but couldn't master new skills without explicit instruction.
AInone | Rating: 53 | 2024-08-20 09:15:35 AM |
OpenAI's latest GPT4-o model has been observed to behave strangely, including copying voices, shouting, and making erotic or violent sounds. This was identified in a 'red teaming' report by OpenAI itself, which aims to identify risks and mitigation strategies. The public version of the model is designed to refuse such behavior and is not expected to suddenly start exhibiting these traits.
AIOpenAI | Rating: 47 | 2024-08-20 09:15:35 AM |
Elon Musk's Grok released a new AI image generation feature, powered by Black Forest Labs, which has few safeguards. The feature allows users to generate fake images, including those of Donald Trump smoking marijuana. The collaboration was announced on Tuesday, and the social media site is already flooded with outrageous images. The feature raises concerns heading into an election cycle.
AIGrok | Rating: 47 | 2024-08-20 09:15:34 AM |
Filmmaking and VFX experts discussed the potential of generative AI and other systems to change the way media is created. The panel consisted of Nikola Todorovic from Wonder Dynamics, Freddy Chavez Olmos from Boxel Studio, and Michael Black from Meshcapade and the Max Planck Institute. They agreed that while AI tools may not be useful in the immediate future, the rate of innovation is rapid and could lead to radical change in the long term.
AItechcrunch | Rating: 47 | 2024-08-20 09:15:34 AM |
World Labs, a startup founded by Stanford University AI professor Fei-Fei Li, has raised $100 million in a funding round led by NEA, valuing the company at over $1 billion. The initial financing in April valued the company at $200 million, with investors including Andreessen Horowitz and Radical Ventures. Fei-Fei Li is a scientific partner at Radical Ventures.
AIWorldLabs | Rating: 57 | 2024-08-20 09:15:34 AM |
Researchers from Cornell, the universities of Washington and Waterloo, and AI2 conducted a study to benchmark hallucinations in generative AI models. They fact-checked models like GPT-4o against authoritative sources on various topics, finding that no model performed exceptionally well across all topics. The study revealed that models that hallucinated the most were those exposed to specific sources of information.
AIAI | Rating: 43 | 2024-08-20 09:15:34 AM |