The Artificial Intelligence Podcast

Only three states enacted laws in 2023 addressing the threat of AI and deepfakes in political campaigns, despite increasing recognition of the potential dangers they pose. There are fears that the upcoming 2024 elections could be marred by disinformation and deepfake videos, leading to calls for more states to enact legal protections. However, obstacles such as reconciling potential regulations with First Amendment rights, the rapid development of AI technology, and enforcement complications have resulted in slow progress.

For the first time ever, the Financial Stability Oversight Council (FSOC) has labeled artificial intelligence (AI) as a potential risk to the stability of the financial system. The US body, formed in response to the 2008 financial crisis, cited potential cybersecurity issues, privacy concerns, and compliance risks. Regulators also highlighted the possibility of AI systems producing biased or inaccurate results, obscuring their own workings, and posing data security, consumer protection and privacy risks. The move follows a similar classification of climate change as an "emerging threat" to the financial stability of the US two years ago.

Intel has unveiled new computer chips including Gaudi3, an AI chip set to compete with Nvidia and AMD. Intel's new chip, designed for generative AI software, is predicted to power demanding AI models. Alongside Gaudi3, Intel announced Core Ultra chips for Windows PCs and laptops, as well as fifth-generation Xeon server chips. All of these products incorporate specialized AI components to enable faster AI program operation.

The Defense Intelligence Agency (DIA) is completing its latest artificial intelligence (AI) strategy, designed to enhance decision support for warfighters and policymakers through advanced human-machine collaboration. Led by DIA's chief technology officer and newly appointed chief artificial intelligence officer, Ramesh Menon, the strategy aims to keep humans in control of technology, rather than vice versa. The plan is organized around five pillars to fully harness AI's potential within the agency, focusing on key areas like platform development, skills improvement, mission priorities, and collaborative partnerships.

As counterfeit products continue to flood the market, potentially costing consumers their safety and $2 trillion annually, companies like Amazon and Entrupy are utilizing AI to combat the issue. Amazon's counterfeit crimes unit uses AI and machine learning to detect potentially illegitimate products, while Entrupy has built a database using machine learning to identify authentic vs counterfeit luxury items. Despite these efforts, the proliferation of fake products, particularly through online platforms, continues to pose a significant challenge.

The European Union has finalized a deal on the world's first comprehensive artificial intelligence (AI) rules. The rules aim for legal oversight of AI technology and address generative AI and police use of facial recognition surveillance. The EU has taken an early lead in crafting global AI regulations, addressing concerns about risk posed by advanced AI systems. The agreement still requires a vote from the European Parliament early next year, but with the deal in place, the vote is considered a formality. The new law wouldn't take full effect until at least 2025.

Join us with Hagay Lupesko, VP of Engineering at MosaicML/Databricks, as we dive into the rapidly evolving world of large language models (LLMs). We'll discuss the latest innovations and challenges in the field, with a spotlight on MosaicML's unique contributions. Additionally, we explore how MosaicML's strategies compare with those of industry giants like OpenAI/Microsoft, Anthropic/AWS, and open-source initiatives such as Meta's LLaMA-2. Hagay provides expert insights into the varied approaches driving AI's future, making this a crucial listen for anyone interested in understanding the trends and potentials of large language models.

Meta has launched "Imagine with Meta AI", a standalone AI image-generation website powered by its Emu image-synthesis model. Trained using 1.1 billion publicly available photos from Facebook and Instagram, the model can create novel images based on text prompts. The service is currently only available in the US. Despite showing average performance compared to other AI synthesis models, concerns are surfacing about potential misuse, reality-warping disinformation, and potentially harmful content.

Meta, OpenAI, and Microsoft have announced they will be purchasing AMD’s latest AI chip, the Instinct MI300X, marking a notable shift from Nvidia's expensive graphics processors essential for AI development. If the MI300X, which is set to ship early next year, performs well and is affordable, it could lower AI development costs and exert competitive pressure on Nvidia. While the pricing for the MI300X hasn't been revealed, AMD CEO Lisa Su hinted it would have to be less than Nvidia's approximated $40,000 price tag per chip.

Google has launched 'Gemini', its largest and most capable artificial intelligence model. Available in three sizes - Ultra, Pro, and Nano, Gemini outperforms OpenAI's GPT-3.5, with plans to license it to customers via Google Cloud. It will also power several Google AI apps, such as the Bard chatbot and Search Generative Experience. Despite lacking details regarding its comparison against GPT-4, the sophisticated AI model seeks to revolutionize advanced customer service, product recommendations, and content creation.