NYTimes: “A start-up founded by three former OpenAI researchers is using the technology development methods behind chatbots to build A.I. technology that can navigate the physical world. Covariant, a robotics company headquartered in Emeryville, Calif., is creating ways for robots to pick up, move and sort items as they are shuttled through warehouses and distribution centers. Its goal is to help robots gain an understanding of what is going on around them and decide what they should do next. The technology also gives robots a broad understanding of the English language, letting people chat with them as if they were chatting with ChatGPT. The technology, still under development, is not perfect. But it is a clear sign that the artificial intelligence systems that drive online chatbots and image generators will also power machines in warehouses, on roadways and in homes.”
WSJ: ““In vivo” gene editing, as the approach is called, could transform medicine. Several of the therapies are for cardiovascular disease, and if proven safe and effective could reach millions of patients…In vivo editing could be less expensive and reach more people than editing cells outside the body. It doesn’t require the laboratories and expertise needed to extract and edit cells. Editing inside the body might also be easier on patients. They don’t have to undergo chemotherapy, for example, which is necessary for sickle-cell patients before receiving their cells that have been edited outside the body. Cells in most of the body can’t be extracted for editing, said Dr. John Leonard, Intellia’s chief executive. “You have to go into the body,” he said.”
FT: “Habits don’t lead to personal optimisation. They lead to suffering…It strikes me that engagement with one’s life, like any relationship, requires ongoing cultivation. It requires effort. The lesson, I think, isn’t that we need pack our days with new experiences, forcing ourselves to swim continually into unfamiliar waters. All the researchers I spoke with stressed the need for balance between the novel and the familiar. It’s that balance I lost and, hopefully, am beginning to find again. It occurs to me that in every language I know, “to be” is an irregular verb.”
Arnold Kling: “I think that the most likely business scenario is not One Model to Rule Them All. Instead, I foresee multiple application-specific uses for machine learning. In medicine alone, there will be applications embedded in robots, applications for aiding in diagnosis, applications to aid in record-keeping, applications to aid in processing insurance claims, and perhaps more. The superpower of LLMs is their ability to communicate with humans in ordinary language. This will make it possible to build applications faster and with a much more intuitive user interface. Under this scenario, an LLM will not require a gigantic base of training data in order to be effective. Instead, it will need training data and human reinforcement that hone its skills in a particular application. A medical diagnosis app does not need to “know” about military history or astrophysics or classical music. It does not have to be trained to give answers that are politically correct. It needs to be trained to give answers that are clinically useful.”
