| Here's your daily AI briefing: - EU countries and lawmakers reach deal on AI regulations.
- Microsoft-OpenAI partnership faces potential U.K. probe.
- Google acknowledges Gemini AI video was edited.
Beth
p/beth-duckett | |
| 1 | | European Union officials reached an agreement Friday on the bloc's rules governing artificial intelligence, known as the AI Act. After days of negotiations, EU delegates reached an agreement on the draft controls, which still require formal approval by member states and Parliament. Exact details of the final agreement were not yet made public on Friday afternoon. More: - On Friday, members of the European Parliament, reps from EU member countries, and European Commission experts agreed on topics like generative AI, facial recognition, and predictive policing.
- For example, facial recognition use by police and governments would be restricted, with some safety and national security exemptions.
- Chatbots and image-manipulation software like "deepfakes" must disclose that their content has been AI-generated, according to the NY Times.
Zoom out: - In prior versions of the draft law, it was noted that violations could lead to fines of up to 40M euros ($43.3M) or 7% of a company's global revenue.
- The rules, which are the first of their kind outside of Asia, will likely serve as a model for similar regulations worldwide.
- Earlier this summer, China finalized certain generative AI rules this summer, which require platform providers offering Gen AI services to the public to obtain licenses to operate. AI-created content is also required to "reflect core socialist values."
|     | |
| 2 | | Microsoft's partnership with OpenAI is potentially facing a full-scale antitrust investigation in the U.K. Britain's competition regulator is seeking feedback about whether to launch the probe, which would determine if the partnership qualifies as a "relevant merger." More: - The agency is taking public comments until Jan. 3 to decide on a formal investigation into whether the collaboration threatens U.K. competition.
- Since 2019, Microsoft has invested as much as $13B into OpenAI, the maker of ChatGPT, and incorporated its AI technology across nearly all of its services.
- The two have remained independent, though Microsoft owns 49% of OpenAI's for-profit operating company.
Zoom out: - In explaining its concerns, Britain's Competition and Markets Authority cited recent developments in OpenAI's governance, some of which involved Microsoft.
- In November, following the firing and hiring of OpenAI's CEO, Microsoft gained a non-voting observer seat on the nonprofit board overseeing OpenAI's for-profit business.
- In a statement, Brad Smith, Microsoft vice chairman and president, said the non-voting observer was "the only thing that has changed" recently in the partnership between the companies.
- This is "very different from an acquisition, such as Google's purchase of DeepMind in the U.K.," he said.
| | | |
| 3 | | A promotional video for Google's new AI model, Gemini, included aspects that were staged, according to an op-ed by Bloomberg. Google has since acknowledged that the video, "Hands-on with Gemini: Interacting with multimodal AI," had edited outputs, and its real voice interaction was edited. More: - Google launched Gemini on Wednesday, saying the advanced AI model is now powering its Bard chatbot.
- A promo video featured spoken user-chatbot conversations and showed Gemini's visual recognition capabilities.
- A description of the video on YouTube notes, "For the purposes of this demo, latency has been reduced, and Gemini outputs have been shortened for brevity."
- Google has since verified to Bloomberg that the demo did not occur in real-time; instead, it used still image frames and fed text prompts to Gemini that the AI model responded to.
- The company told CNBC that the video was "an illustrative depiction of the possibilities of interacting with Gemini, based on real multimodal prompts and outputs from testing."
| | | |
| 4 | | AI apps and websites that manipulate images to "undress women" have experienced a surge in popularity, with 24 million visitors in September, according to a Graphika report. The sites and apps allow users to upload images of fully clothed individuals, which are then processed by AI to create fake nude images of them. More: - According to Graphika, there was an over 2,400% surge in referral link spam for the service on sites like X and Reddit since the start of the year,
- Additionally, the report identified 53 Telegram groups with at least 1 million users that are known to access the services.
- The report warned that the service could result in more online harm, including harassment, sextortion, and non-consensual image sharing.
Zoom out: - The apps' popularity is tied to the rollout of more advanced open-source AI models that create superior images, and their free availability adds to their allure, noted Santiago Lakatos, a Graphika analyst.
| | | |
| 5 | | Presto, a provider of AI voice ordering at drive-thrus, relies on off-site human workers to help fulfill over 70% of orders, according to SEC filings. The SEC opened an investigation into Presto Automation's AI technology this year, according to Bloomberg. Details: Presto develops voice AI technology mainly for restaurant drive-thrus, including Hardee's, Del Taco, and Checkers. While it markets its technology as an AI-driven process, the company also relies on human workers, mainly from the Philippines, for over 70% of customer interactions to ensure accurate orders. Quotable: "Our human agents enter, review, validate, and correct orders," CEO Xavier Casanova said during an earnings call. "Human agents will always play a role in ensuring order accuracy." What it means: In this case, labor replacement isn't linked to AI but rather to outsourcing. Presto notes that it expects to rely less on human intervention as its AI voice assistant improves over time. What the numbers show: While fast-food chains and restaurants are relying more on AI and automation in areas like order and staff management, a Bernstein analysis shows only about 9% of the industry has adopted drive-thru automation. 
| | | |
| 6 | | Weekly roundup — The top AI stories you might have missed this week: - Google's Bard chatbot is now being powered by Gemini, the company's most advanced AI model. According to Google, it's the biggest improvement to Bard since its launch in March. CEO Sundar Pichai said the model will eventually come to Google's search engine, ad products, Chrome browser, and beyond.
- Advanced Micro Devices (AMD) is now selling its MI300X chip, an AI accelerator that competes against Nvidia's H100 chip for AI training and inference. Meta, Microsoft, Oracle, and OpenAI have already signed up to use the MI300X when it starts shipping next year, taking on market leader Nvidia.
- Meta's new web-based AI image generator is powered by its Emu model, which was trained on 1.1 billion publicly accessible Facebook and Instagram images. Users of Facebook and Instagram users may have unknowingly contributed to Emu's training data with their photos, prompting some concerns about how Meta uses that data.
- Meta and IBM are leading a new alliance of companies, universities, and organizations focused on open-source AI development. The AI Alliance counts more than 50 members, including Intel, Oracle, Dell, AMD, Sony, Hugging Face, UC Berkeley, Yale, NASA, Cleveland Clinic, CERN, and others.
- ChatGPT was the most viewed article on English Wikipedia in 2023, according to the Wikimedia Foundation. The article on OpenAI's chatbot dominated the list with nearly 49.5 million page views this year.
| | | |
| 7 | | QUICK HITS - Want to master Kubernetes? The Kubernetes Cookbook is a must-have guide to discover guidance on cloud architecture analytics, optimization, governance, and more.*
- X has rolled out Grok, its "rebellious" AI chatbot from Elon Musk's xAI startup, to its premium+ subscribers in the U.S. Musk said the bot will expand to all English language users in about a week.
- PwC US is launching ChatPwC, its in-house AI tool powered by OpenAI, to all of its 75,000 employees by year-end.
- Google's AI-driven note-taking app, NotebookLM, has launched in the U.S. The app can summarize imported documents, generate key points, answer questions about sources, and change notes into different document formats.
- Toast to 2024: Key affiliate marketing trends to help your campaigns start off on the right foot in the new year. Get the free report.*
*This is a sponsored post. | | | |
 | AI and technology writer | | Beth is a contributing editor and writer of Inside's AI and Tech newsletters. She has written for notable publications like USA Today and the Arizona Business Gazette. During her time as a public policy reporter at The Arizona Republic, she received recognition with a Pulitzer Prize nomination and a First Amendment Award for her collaborative reporting on pension cost increases. Beth also authored a book on the solar photovoltaic industry in 2016. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
| |
Comments
Post a Comment