September 18, 2023
-
Research
Multi-AI collaboration helps reasoning and factual accuracy in large language models
MIT, 09/18/23. In a world where artificial intelligence (AI) is increasingly relied upon, a team from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has found that collaboration among multiple language models can lead to more accurate and comprehensive solutions. By engaging in discourse and debate, these AI models sharpen their own answers and improve their problem-solving abilities. This approach not only enhances the performance of existing language models but also offers an automatic means of self-improvement. The method has shown promising results in mathematical problem-solving and has the potential to be applied across various domains and modalities. READ THE ARTICLE
-
Regulation
Harsh AI Regulation From Congress Imperils Innovation
Reason, 09/18/23. Calls for a new agency to regulate AI and a licensing regime have been growing in Congress. However, these proposals could hinder American innovation and favor larger companies. Past licensing regimes have shown to be influenced by existing players, resulting in higher costs and excluding new and innovative players. Additionally, an AI regulator could interfere in every sector of the economy, raising concerns about the administrative state and agency capture. Instead of adding more regulation, policymakers should focus on existing laws and consider repealing or reforming statutes to address AI concerns. READ THE ARTICLE
-
Jobs
Inside PwC’s push to train its workers on AI
Business Insider, 09/18/23. PricewaterhouseCoopers (PwC) is taking steps to integrate artificial intelligence (AI) into its operations. The company plans to provide AI training to its 75,000 employees in the US and Mexico, enabling them to incorporate AI into their work and support clients in AI transformations. This move highlights the growing importance of AI in the business world, with PwC investing $1 billion over three years in expanding its AI offerings. The training will be conducted in monthly batches, with lessons delivered online, and will involve AI experts sharing their knowledge with colleagues. PwC aims to have all employees knowledgeable and capable of utilizing AI effectively. READ THE ARTICLE
-
OpenAI
ChatGPT masquerades as AI but is something much different, researcher says
The Badger Herald, 09/18/23. In her lecture, linguistics professor Emily Bender discussed the safety, appropriateness, and desirability of synthetic text, specifically focusing on large language models like ChatGPT. She explained that these models have evolved from corpus models to neural nets, which can understand similarities in word meanings. However, large language models still lack a true understanding of meaning. Bender outlined the issues associated with these models, including exploitative labor practices, biased training data, data theft, and a lack of accountability. To responsibly use such models, Bender emphasized the need for labor rights, regulation, transparency, and accountability. Ethical applications of large language models include language learning partners, video game characters, and writing support. In conclusion, society should strive for safer and more accountable use of synthetic text. READ THE ARTICLE
-
Healthcare
Urgent Call for AI to “Do No Harm” in Biomedicine
Stanford HAI, 09/18/23. In a recent paper published in Science Translational Medicine, Stanford University fellows Carole Federico and Artem Trotsyuk argue for the need to ensure the safety and equity of AI in hospitals. They emphasize the importance of implementing changes today to protect patients. The authors propose a top-down approach with regulatory coordination among federal agencies, healthcare institutions, and review boards. They also suggest guidelines for research oversight, post-approval monitoring, transparency in AI design, and patient opt-out options. By taking immediate action and establishing clear standards, the authors believe AI can be used beneficially in healthcare without causing harm. READ THE ARTICLE
-
Adoption
You’re Out of Time to Wait and See on AI
Bain & Co., 09/18/23. The rapid advancement of artificial intelligence (AI) has led to a technology revolution that is transforming various industries. General purpose foundation models, such as ChatGPT, have bridged the gap between human intelligence and machine intelligence, gaining over 100 million users within two months. This revolution is not only a technological one but also an economic one, as the cost of building foundation models has decreased, allowing companies to experiment with AI and generative AI at a relatively low cost. Companies are already seeing productivity gains and cost reductions in areas like sales, marketing, HR, and software development. Moreover, foundation models and generative AI are enabling new value creation in areas like rapid upskilling, hyper-personalization, redefining the customer experience, and core product innovation. To fully leverage the potential of AI, tech leaders need to address technical and process aspects, such as privacy and security, machine learning operations, scalability, and change management. Waiting to see what competitors do is not an option, as AI is quickly evolving, and companies need to seize the opportunities it presents to create new value. READ THE ARTICLE
-
Google
Google Bard AI won’t answer questions about Putin asked in Russian
NewScientist, 09/18/23. Scientists are making exciting discoveries by cracking historical codes to reveal lost secrets. This process involves decoding ancient scripts and languages to unlock hidden knowledge from the past. In another breakthrough, the world’s brightest X-ray machine has been turned on, providing scientists with a powerful tool for studying and understanding the molecular structure of materials. Additionally, researchers have determined the exact number of cells in the human body, shedding light on our biological composition. However, not all scientific phenomena go as expected, as a NASA spacecraft finds that an asteroid behaves unexpectedly. CBD, a popular substance, is also under scrutiny, with scientists examining whether it is a wonder drug or a waste of money. Artificial intelligence has its limitations too, with Google’s Bard AI unable to answer questions about Putin in Russian. In the realm of medical advancements, an implant now allows people to type on virtual keyboards using only brain signals. However, doctors are concerned about the growing number of people with obsolete medical implants. In the field of astronomy, the James Webb Space Telescope (JWST) captures a stunning image of a baby star with powerful jets, deepening our understanding of stellar formation. Lastly, unravelling the secrets of the vagus nerve holds promise for revolutionizing medicine. These recent scientific discoveries and challenges provide a glimpse into the ever-changing and fascinating world of scientific exploration. READ THE ARTICLE