AI Update: Time to Start Getting Superaligned

Welcome to the latest edition of FindBiometrics AI update. Here’s the latest big news on the shifting landscape of AI and identity technology:

OpenAI, the company behind ChatGPT, has announced a new initiative aimed at ensuring that AI doesn’t wipe out humanity. AI researchers call the effort to build ethical guardrails around the technology “alignment” — the task of getting artificial intelligence aligned with human goals and values. OpenAI’s initiative is dubbed “Superalignment“, and is aimed at getting ahead of not just the expected emergence of “artificial general intelligence” (AGI) but also the potential “superintelligence” that could arise from AI technology that is vastly more sophisticated than the human brain. OpenAI says it will dedicate 20 percent of its current computing power to this project, and is looking for machine learning researchers and engineers who want to help.

Silicon Valley startup Deepmedia is in talks with multiple Asian governments and media corporations concerning its deepfake detection technology, according to CEO Rijul Gupta. Having recently signed a $25 million, three-year contract with the US Department of Defence, the company is now looking to help entities in the APAC region to flag AI-generated videos and speech, often with political interference as a key concern. Gupta says his firm’s technology has a 99 percent success rate. It can also produce deepfakes. “We need just five seconds of your voice, and we can make you say anything in any language with near-perfect accuracy,” he told Nikkei Asia.

China has announced new controls on the export of gallium and germanium, metals that are used in high-end semiconductors. This could make work difficult for chipmakers elsewhere; China accounts for about 80 percent of the world’s supply in such metals, and may be responsible for half of America’s germanium supply. The controls will come into effect on August 1, and will require domestic exporters to get new licenses from the Chinese government.

Nvidia has acquired OmniML, an AI startup that developed software designed to shrink machine learning models to run on a device, rather than in the cloud. It did so quietly, back in February, about a month after OmniML announced a strategic partnership with Intel. The acquisition suggests that Nvidia is looking to expand its market presence through the production of AI chips designed for specific things like drones and robots.

India’s electronics and communications minister, Ashwini Vaishnaw, says US-based Micron Technology will break ground on the country’s first semiconductor assembly plant next month. The $2.75 billion project is expected to result in the first domestically-made microchips by the end of next year. The government is also considering bids for a $10 billion subsidy program for chipmakers, asking for proposals to produce “mature nodes” of at least 40 nanometers in size.

Louisiana has become one of the first US states to criminalize certain uses of AI-generated deepfake technology. It is now illegal in the state to create or even possess a deepfake that knowingly depicts a minor engaging in sexual conduct, with potential penalties of a five to 20 years in prison and a $10,000 fine. It also criminalizes the distribution of deepfakes depicting non-consenting adults in sexual acts. So far nine states have enacted laws regulating deepfakes: Washington, California, Wyoming, Minnesota, Texas, Georgia, Virginia, and New York, in addition to Louisiana.

NEC Corporation has announced its own “NEC Generative AI Service”, through which it will offer licenses for the use of proprietary large language models, as well as related hardware, software, and services. In the identity space, the company is best known for its sophisticated NeoFace facial recognition solution; but its generative AI platform is focused on language-based systems, and not the generation of images such as deepfakes. The Tokyo-headquartered conglomerate says its LLM is uniquely proficient in Japanese, and that it has been using it internally for coding and document creation. NEC has also launched an “NEC Generative AI Hub” business unit aimed at guiding clients through their own integration of AI technology.

Israel-based Silicom Ltd. has partnered with Hailo to integrate its AI-focused processors into a new product platform dubbed “Edge AI”. Silicom specializes in networking and data infrastructure solutions; the company says this will be its first foray into artificial intelligence. No specific Edge AI products have yet been announced, but Silicom CEO Liron Eizenman says his firm’s collaboration with Hailo will make it “cost-effective to move behavior analytics, human intrusion detection, facial recognition and vehicle analytics to the edge.” The news comes after Hailo worked with Taiwan-based QNAP Systems on an AI-driven surveillance system earlier this year.

Meanwhile, Pope Francis has partnered with Santa Clara University’s Markkula Center for Applied Ethics to form a new body called the Institute for Technology, Ethics, and Culture (ITEC), which has published a new handbook entitled “Ethics in the Age of Disruptive Technologies: An Operational Roadmap“. The handbook is aimed at offering some ethical guidance in areas of technology including artificial intelligence, machine learning, encryption, and even facial recognition.

The Chatbot’s Take:

–

July 7, 2023 – by Alex Perala