The Artificial Intelligence Podcast

Google DeepMind has unveiled SIMA, an AI agent designed to play video games as a human rather than a dominant AI. SIMA, which stands for Scalable, Instructable, Multiworld Agent, is still in the research phase but aims to eventually play any video game, including open-world games. The purpose of SIMA is not to replace game AI but to act as another player that integrates well with the player's party. It combines natural language instruction with 3D world understanding and image recognition. SIMA has been trained using games like No Man's Sky and Goat Simulator 3, and currently possesses around 600 basic skills.

Walmart is expanding its business activities by selling its AI software, which was previously used internally, to other companies. The software uses artificial intelligence to optimize the routes for truck deliveries, leading to savings of $90 million for Walmart in a year. The company aims to empower other businesses with the same successful technologies that it has utilized. The pricing of the software will depend on factors such as usage volume and frequency. This move is part of Walmart's strategy to diversify its business beyond retail, following the trend of retailers entering faster-growing areas with higher profit margins. Other retailers, including Target and Kroger, have also been diversifying their businesses in a similar way.

Midjourney, a popular AI image-generator, has announced that it will block users from creating fake images of President Joe Biden and former President Donald Trump ahead of the U.S. presidential election. The decision is a temporary measure to prevent misuse of the tool during the election season. CEO David Holz emphasized that Midjourney is focused on exploring new mediums of thought and expanding imaginative powers rather than engaging in political speech. However, a watchdog group report revealed that Midjourney had performed poorly in generating images supporting disinformation and false claims of election fraud. The company's decision to block images of Biden and Trump aims to address this issue.

More and more businesses are turning to artificial intelligence (AI) models to perform tasks previously done by humans. However, AI has a tendency to generate false responses, known as "hallucination," which makes it unlikely that AI will completely replace human workers. Despite the use of AI, human workers will still be necessary to supervise and ensure the accuracy and quality of AI outputs. The introduction of AI into the workplace also raises questions about how it affects employees' exemption status under the Fair Labor Standards Act (FLSA). Job duties impacted by the use of AI tools may change employees' eligibility for overtime exemptions, requiring employers to regularly assess and adjust job duties to avoid misclassification claims.

Artificial intelligence (AI) is being used to create and spread disinformation during elections, posing a significant threat to democratic processes worldwide. With the help of AI technologies, anyone can now create convincing "deepfakes" with a simple text prompt. Governments and organizations have started responding to this threat, with the Federal Communications Commission outlawing AI-generated robocalls aimed at discouraging voters in the US, and major tech companies signing an agreement to prevent the use of AI in disrupting elections globally. However, identifying those responsible for creating AI deepfakes is difficult, and governments and companies are struggling to keep up with the deluge of fake content. Efforts to regulate AI deepfakes must also be careful to avoid unintended consequences and protect free speech.

Generative AI, which can create original content like text, video, and images, is susceptible to data poisoning. Hackers can insert false or misleading information into the data used to train AI models, leading to the spread of misinformation. Generative AI models rely on data from the open web, making it easy for hackers to manipulate. Even a small amount of false information can significantly impact the outputs of AI models. Researchers warn that this poses a risk of disseminating harmful information or unknowingly sharing sensitive data. Legislation, improved safety measures, and increased awareness are needed to address this issue and ensure responsible use of generative AI.

Members of the U.S. Senate Rules Committee have expressed concerns about the potential threats to fall elections, particularly regarding the use of artificial intelligence (AI) to spread disinformation and intimidate election workers. Senators highlighted the dangers of AI in manipulating election outcomes, including the use of deepfake videos and ads to attack candidates. They also raised worries about the security of future elections and emphasized the role of AI and social media in spreading misinformation. Election officials discussed the threats faced by election workers, including the safety of poll workers, and called for federal legislation to protect them. The hearing also touched on the ongoing debate over voting rights.

Scientists in the field of artificial intelligence (AI) are urging for the responsible and ethical use of protein design tools to prevent misuse as bioweapons. To address these concerns and avoid excessive government regulation, researchers have launched an initiative to promote responsible protein design. The initiative has gained support from scientists who utilize AI in biological design. However, some experts believe that government action and regulations should accompany the voluntary guidance. Reports have highlighted the potential for AI tools to aid in the development of biological weapons, raising fears of creating dangerous toxins or highly contagious viruses. The newly launched initiative calls for self-regulation within the biodesign community and emphasizes the need for enhanced DNA synthesis screening. Governments are also considering biosecurity risks associated with AI, including the possibility of introducing screening for federally funded research. While some hope to avoid government regulation for continued advancement, others believe a combination of self-regulation and government oversight is necessary.

California is advancing regulations on the use of artificial intelligence (AI) by businesses, with the aim of protecting personal privacy. The rules, which apply to companies with annual revenues exceeding $25 million or processing data from over 100,000 Californians, require businesses to notify individuals before using AI and prohibit discrimination against those who opt out of interacting with AI models. The regulations also demand risk assessments by employers or contractors to evaluate the performance of AI technology. However, organizations have voiced concerns about potential loopholes in the rules that could allow companies to evade accountability. The final vote on the rules is expected to take place in about a year.

San Francisco-based data cloud vendor, Databricks, has partnered with generative AI specialist, Mistral AI, and made an investment in the company. The move follows similar partnerships formed by Microsoft and Snowflake with Mistral AI. As part of the partnership, Mistral AI's open source language models will be integrated into Databricks' Data Intelligence Platform, providing customers with more options and flexibility in generative AI development. Databricks already offers access to language models from various vendors and aims to expand its model-building and fine-tuning options in the future. The company continues to prioritize AI and stay competitive in the evolving generative AI landscape.