In July, AI technology supported one of the longest-standing world traditions – the Olympics, where Google secured the position of AI sponsor of the US team. But France proved notorious for not only hosting the Olympic games.
French authorities are setting their sights on Nvidia, a chipmaker vital for AI development. Also, OpenAI and Meta haven’t lost velocity in their race for AI chatbot market dominance.
07.01.2024 Apple introduces 4M model
Apple has released a public demo of their 4M AI model. The system was developed together with the Federal Institute of Technology Lausanne (EPFL) of Switzerland. The solution is capable of generation across multiple modalities, so users can ask it to create images from descriptions, detect objects, and manipulate 3D scenes using text prompts.
More can be found in VentureBeat.
07.01.2024 AI writer fails at Wimbledon
Wimbledon’s Catch Me Up online feature is a generative AI-powered service, complimentary to human-delivered reports. The problem lay with the system confusing key facts about players and matches. For example, it described former US Open champion Emma Raducanu as British No. 1, although she is No. 3. And it was only one of many of the system’s mistakes, as spotted by X (formedly Twitter) users.
More information can be found in The Guardian.
07.02.2024 Google’s emissions climb by nearly 50% due to AI
Google has revealed that its greenhouse gas emissions have risen by 48% over the last five years. The rise is driven mostly by the development of AI products. The International Energy Agency estimates that global data centers will consume 1000TWh (terawatt hours) in 2026, matching Japan’s national electricity demand. By 2030, global AI energy consumption is expected to reach 4.5% of total global energy consumption.
More can be found in The Guardian.
07.03.2024 Apple gets “observer” seat on OpenAI board
After the announcement of the Apple-OpenAI partnership, the company has secured a seat on the OpenAI board in the same way Microsoft has done. Apple will bring OpenAI integration to iOS 18 as a part of their new features.
More can be found in Bloomberg.
07.04.2024 WhatsApp develops personalized AI avatars
WhatsApp is developing a generative AI feature to create personalized avatars for users in various settings. Found in the new WhatsApp Beta for Android 2.24.14.7 by WABetaInfo, the feature uses user images, text prompts, and Meta’s AI Llama model to generate these avatars.
More can be found in TheVerge.
07.04.2024 French Antitrust office targets Nvidia
The French antitrust office prepares to charge Nvidia, currently the world’s most valuable company, with anti-competitive practices. According to a Reuter’s report, the world’s largest AI chip manufacturer has leveraged its position to secure dominance – particularly in CUDA chip programming software, which is currently the only software that is 100% compatible with the GPUs used in AI model development.
More can be found in the Reuter’s report.
07.09.2024 Department of Justice seizes Russian bot farm
The Department of Justice announced the seizure of two domain names and over 900 social media accounts linked to an AI-enhanced Russian bot farm. These accounts, many mimicking American identities, posted content supporting Russia’s actions in the Russia-Ukraine war. The DOJ claims an employee of RT, Russia’s state media outlet, with leadership approval, used the bot farm to amplify RT’s social media presence. The employee allegedly bought domain names from Namecheap to create email servers, which generated 968 email addresses for social media accounts. In early 2023, a member of FSB, Russia’s Federal Security Service, formed a private intelligence group including RT employees intended to spread disinformation via the bot farm, advancing the FSB’s mission, according to the DOJ.
More can be found in The Verge.
07.09.2024 Washington Post introduces Climate Answers chatbot
The Washington Post is introducing an AI chatbot called Climate Answers to its homepage, app and articles. This tool uses the outlet’s reports to answer questions about climate change, the environment, and sustainable energy. Users can ask questions like, “Should I get solar panels for my home?” or “Where in the US are sea levels rising the fastest?” and the chatbot provides summaries based on The Washington Post’s climate articles dating back to 2016. Answers include links to the source articles and relevant excerpts. The tool is based on a large language model from OpenAI, with experiments involving AI models from Mistral and Meta’s Llama.
More can be found in The Verge.
07.10.2024 Microsoft and Apple will not take board seats at OpenAI
Microsoft and Apple have decided against taking board seats at OpenAI due to increased regulatory scrutiny. Microsoft withdrew from the OpenAI board a year after its $13 billion investment, citing confidence in OpenAI’s progress. Contrary to reports, Apple will also not join the board. OpenAI expressed gratitude to Microsoft and looks forward to continuing their partnership. This decision comes amid growing regulatory concerns about big tech’s influence on AI development.
More can be found in AINews.
07.12.2024 AI Act countdown starts
Tech companies that want to offer AI-powered services in the EU will have time until August 2nd 2026 to adjust their offers to meet the new regulations in the European Union. The new law regulates the rules regarding AI in relations to copyright, system transparency, and impact on human lives. The law was officially passed in March after two years of work.
More can be found in TheVerge.
07.12.2024 OpenAI introduced a system to assess AI power
The company uses a 1 to 5 scale to determine the power of an AI system. The scale was designed to track progress toward developing Artificial General Intelligence. Basically, the scale starts with Level 1, where modern LLMs like the one used in Chat GPT currently stand. Level 2 is defined as a model capable of solving problems at the level of a person with a PhD. Level 3 is used for agents that can take actions on the user’s behalf. Level 4 is where the AI participates in the creation of innovations. Last, 5 level, is the AGI, in which AI can perform the work of an entire organization.
More can be found in TheVerge.
07.15.2024 Google tests Google Vids video presentation creator
Google Vids is supposed to create video presentations using prompts, uploaded documents, images, and videos provided by the user. Despite the apparent similarity to OpenAI’s Sora’s video generation capabilities, the system is different, designed to create business video presentations from delivered materials. The system is also supported by Gemini’s text generation capabilities.
More can be found in TheVerge.
07.17.2024 Over 40% of Japanese companies don’t have plans to use AI
According to a survey conducted by Reuters, 24% of Japanese companies already use AI in their operations and a further 35% are planning to do so. Yet the survey revealed that 41% of companies have no plans on using AI in their daily operations. The survey was conducted across 506 companies, with roughly 250 choosing to respond.
More can be found in Reuters.
07.18.2024 Meta will not release its multimodal Llama mode in the EU
Meta will not launch its upcoming multimodal AI model in the European Union due to regulatory concerns, preventing European companies from using it despite its open license release. Spokesperson Kate McLaughlin cited the unpredictable European regulatory environment. The EU recently set compliance deadlines under its new AI Act, with tech companies having until August 2026 to comply. Meta’s decision follows a similar move by Apple regarding its Apple Intelligence rollout due to the Digital Markets Act. Meta has also halted plans to release its AI assistant in the EU and paused generative AI tools in Brazil over data protection compliance concerns.
More can be found in TheVerge.
07.18.2024 GPT-4o mini is a smaller and cheaper multimodal AI model
According to OpenAI’s announcement, the model is the most cost-efficient tool currently available, costing $0.15 per 1 million tokens of user input and $0.60 for every million received. It is far cheaper than the full-scale GPT-4o, which costs $5 per 1 million input tokens and $15 for one million output tokens.
More can be found in the OpenAI blog.
07.18.2024 Groq releases open-source models that outperform proprietary ones
Groq, an AI hardware startup, has released two open-source language models that outperform leading proprietary models in specialized tool use capabilities. The Llama-3-Groq-70B-Tool-Use model has taken top position on the Berkeley Function Calling Leaderboard, surpassing models from OpenAI, Google, and Anthropic. Rick Lamers, project lead at Groq, announced this achievement on X (formerly Twitter), highlighting that their models, including the Llama 3 Groq Tool Use 8B and 70B, have outperformed the competition, like Claude Sonnet 3.5, GPT-4 Turbo, GPT-4o, and Gemini 1.5 Pro.
More can be found in VentureBeat.
07.19.2024 Google becomes “official AI Sponsor” for the US Olympic team
The tech giant will support the US Olympic team and launch a number of features that will enrich the viewer experience of the event. The broadcast will be covered by NBCUniversal and Google will enrich it with Google Maps’ 3D views of Paris. Also, the commentators will use Google Gemini model’s analysis and briefings in their work to answer Olympic and Paralympic questions.
More can be found in TheVerge.
07.23.2024 Meta launches the Llama 3.1 405B-based open source AI app
The company behind Facebook and Instagram has launched an app aimed to rival the products of OpenAI and Anthropic. According to the company’s statement, the open-source LLM is competitive and will provide users with a similar experience compared to services currently available.
More can be found in The Guardian.
07.24.2024 Reddit blocks non-paying search engines
Google is the only mainstream search engine that will show results from Reddit. The company started to block search engines and crawlers from accessing their service content. The block also aims to prevent AI companies from using its content for training Large Language Models. Reddit remains visible in Google due to a recent $60 million deal with Google that permits the company to develop LLMs using Reddit posts and interactions.
More can be found in TheVerge.
07.24.2024 Mistral releases Mistral Large 2
The 123 billion parameter model will open for research and non-commercial uses. The model will be released with weights, allowing third parties to fine-tune it to adjust to their needs. For companies wishing to use the model for commercial purposes, a separate license will be required.
More can be found in VentureBeat.
07.24.2024 StackOverflow: Developers not concerned about AI overtaking their jobs
According to a survey by StackOverflow, an international community of developers and programming knowledge-sharing platform, generative AI technology is not considered a threat from the software developers’ perspective. The survey also shows that 76% of developers use AI tools in their jobs. Also, 81% of developers see increased productivity from the use of AI tools.
More can be found on VentureBeat.
