Loading summary
Transcript2 lines
- [00:04]
Brian McCullough
Welcome to the TechMe Bright Home for Friday, December 20, 2024. I'm Brian McCullough. Today, Gemini Flash Thinking is a model making a serious run at reasoning. The FAA has had to ban drones in New Jersey because of all those weird drone sightings. Self driving technology continues to prove it is safer, an open source generative physics engine for robots and of course the weekend long Read Suggestions here's what you missed today in the world of tech. Google has released Gemini 2.0 flash thinking, an experimental reasoning model that explicitly shows its thoughts and can use them to strengthen its reasoning. Quoting TechCrunch the new model called Gemini 2.0 flash thinking experimental, a mouthful to be sure, is available in AI Studio, Google's AI prototyping platform. A model card describes it as best for multimodal understanding, understanding, reasoning and coding with the ability to reason over the most complex problems in fields such as programming, math and physics. In a post on X, Logan Kilpatrick, who leads product for AI studio, called Gemini 2.0 flash thinking experimental the first step in Google's reasoning journey. Jeff Dean, chief scientist for Google DeepMind, Google's AI research division, said in his own blog post that Gemini 2.0 flash thinking experimental is trained to use thoughts to strengthen its reasoning. We see promising results when we increase inference time computation, dean said, referring to the amount of computation used to run the model as it considers a question end quoting the Verge. The demo shared by Dean shows how Gemini 2.0 flash thinking goes about answering a physics problem by thinking through a series of steps before offering a solution. This isn't necessarily reasoning in the way humans perform it, but it means the machine breaks down instructions into smaller tasks that can produce stronger outcomes. Another example posted by Google product lead Logan Kilpatrick shows the model reasoning its way through a problem that involves both visual and textual elements. You can try Gemini 2.0 flash thinking on Google's AI Studio. End quote and quoting VentureBeat Unlike competitor reasoning models 01 and 01, many from OpenAI, Gemini 2.0 enables users to access its step by step reasoning through a drop down menu, offering clearer, more transparent insight into how the model arrives at its conclusions by allowing users to see how decisions are made. Gemini 2.0 2.0 addresses long standing concerns about AI functioning as a black box and brings this model, licensing terms still unclear, to parody with other open source models fielded by competitors. My early simple tests of the model showed it correctly and speedily within one to three seconds answered some questions that have been notoriously tricky for other AI models, such as counting the number of Rs in the word strawberry. In another test, when comparing two decimal numbers, 9.9 and 9.11, the model systematically broke down the problem into smaller step from analyzing whole numbers to comparing decimal places. These results are backed up by independent third party analysis from LM arena, which named Gemini 2.0 Flash Thinking the number one performing model across all LLM categories. Gemini 2.0 Flash Thinking's multimodal capabilities expands its potential use cases, enabling it to tackle scenarios that combine different types of data. For example, in one test, the model solved a puzzle that required analyzing textual and visual elements, demonstrating its versatility in integrating and reason across formats. Developers can leverage these features via Google AI studio and Vertex AI, where the model is available for experimentation. Quoting DD DDoS on X Google really cooked with Gemini 2.0 flash thinking. It thinks and it's fast and it's high quality. Not only is it number one on LM arena on every category, but it crushes my go to Math riddle in 14 seconds, five times faster than any other model can solve it. Google is making OpenAI dance end quote the US Federal Aviation Administration is prohibiting drone flights over several parts of New Jersey as public hysteria over sightings of possible drone activity continues to persist. Quoting the New York Times the Federal Aviation Administration has issued a month long ban on drone flights over large swaths of New Jersey, the first broad prohibition of its kind since the authorities began investigating a spate of sightings last month that set off fear and speculation. The ban began late on Wednesday and will continue through January 17, according to an FAA alert. The notification cited special security reasons for prohibiting flights in airspace near 22 New Jersey communities, including three of the state's largest cities Camden, Elizabeth and Jersey City. Only drone pilots authorized to operate for national defense, law enforcement or disaster response purposes are permitted to send the unmanned crafts. Alo operators of drones used for commercial purposes may apply for a waiver after providing a, quote, valid statement of work, the FAA alert said. The FAA said in a statement that it had temporarily restricted drone flights over critical New Jersey infrastructure at the request of what it described as federal security partners. Officials referred additional questions to the U.S. department of Homeland Security, end quote quoting Bloomberg. The agency has published 22 temporary flight restrictions over critical infrastructure in the state, according to a statement on Thursday, adding that it was instit the band at the request of federal security partners. The announcement comes a day after the FAA warned people that pointing lasers at aircraft is illegal and can result in a hefty fine after seeing a dramatic increase in pilots reporting such incidents so far this month. End quote and quoting NBC News for about the past month, drones have reportedly been lighting up the night skies with no real explanation of who is operating them and the purpose of the flights. There has been, though, creeping skepticism about whether most of the reports are due to people mist mistaking stars or the routine flights of aircraft for drones. Alarmed New Jersey residents and local politicians have been demanding information about the nightly hovering objects. No person, group or government agency has taken responsibility for them, end quote. Soon, you might be able to change your entire outfit on an Instagram post, thanks to AI Instagram head Adam Mosseri has teased an AI editing tool in Instagram powered by Meta's moviegen AI model that will allow you to change outfits, backgrounds and more in the videos you post via a text prompt quoting the Verge. Instagram is planning to introduce a generative AI editing feature next year that will allow users to, quote, change nearly any aspect of your videos. The tech is powered by Meta's moviegen AI model, according to a teaser posted by Instagram head Adam Mosseri, and aims to provide creators with more tools to help transform their content and bring their ideas to life. Without extensive video editing or manipulation skills, Mosseri says the feature can make adjustments using a simple text prompt. The announcement video includes previews of early research AI models that change Mosseri's outfit, background, environments, and even his overall appearance in one scene, transforming him into a felt puppet. Other changes are more subtle, such as adding new objects to the existing background or a gold chain around Mosseri's neck without altering the rest of his clothing. In the preview, Mosseri demonstrated remarkable stable background and clothing modifications that maintain their integrity even during movements, though we're only seeing fleeting movements in action. While these clips echo the initial polish we saw with OpenAI's SORA demonstrations, it is worth noting that SORA's public release hasn't quite matched those early expectations. The true capabilities of Instagram's AI video tools will only become clear once they're launched. As mentioned, the technology builds on Meta's moviegen AI video platform unveiled back in October, which emphasizes its ability to maintain human elements and motion fidelity in both generated and edited content. While competitors like OpenAI, Sora and Adobe's Firefly video model moved earlier in the space. With Adobe already integrating text to video features into Premiere Pro's beta, Meta is taking a more measured approach, while moviegen's broader release timeline remains undefined. Instagram marks its first confirmed implementation of this text to video technology. If you listen to this podcast, you care about what's going on in the world. Listening to us is great, but if you want even more great coverage of our crazy world, you should subscribe to the Washington Post. You don't really need me to tell you about the Washington Post when it comes to their tech coverage because I quote from them all the time. But it's not just tech that they're good at, they're one of my personal go to's for all the things beyond tech too. I even signed up to get the Post's for you newsletter, which sends me my very own personalized roundup of stories every morning based on my interests and reading history. Their app makes it easy for me to stay up to date on the latest news, save and share stories, and follow my favorite authors. The Post even offers a cool feature for audio lovers like you. You can actually listen to articles in addition to reading them so you can tackle your to do list and catch up on the news at the same time. Now more than ever, it's important to stay up to date on the world, so go to washington post.com Ride to subscribe for just 50 cents per week for your first year. That's 80% off their typical offer, so this is truly a steal. Once again, that's washingtonpost.com Ride to subscribe for 50 cents per week for your first year. Lumen is the world's first handheld metabolic coach. It's a device that measures your metabolism through your breath, and on the app, it lets you know if you're burning fat or carbs and gives you tailored guidance to improve your nutrition, your workouts, your sleep, even your stress management. All you have to do is breathe into your Lumen first thing in the morning and you'll know what's going on with your metabolism then. Lumen gives you a personalized nutrition plan for that day based on your measurements. You can also breathe into it before and after workouts and meals so you know exactly what's going on in your body in real time. And Lumen will give you tips to keep you on top of your health game. Because your metabolism is at the center of everything your body does, optimal metabolic health translates to a bunch of benefits, including easier weight management, improved energy levels, better fitness results, better sleep, etc. So if you want to stay on track with your health this holiday season, go to Lumen m to get 15% off your lumen. That's L U M E N me ride for 15% off your purchase. Lumen makes a great gift too. Thank you Lumen for sponsoring this episode. All right, I'm throwing in the towel. Clearly my voice is deteriorating. It's actually getting painful at this point. So here you go. Rest of the show will be robot voice.
- [11:16]
Logan Kilpatrick
Brian, Remember that one of the long standing promises of self driving technology was that it could theoretically be safer, much safer than human drivers. How is that going? Well, pretty good actually. Quoting the Verge Waymo's autonomous vehicles cause less property damage and fewer bodily injuries when they crash than human driven vehicles, according to a study that relies on an analysis of insurance data. The study is the product of the collaboration between Waymo and insurer Swiss Re, which analyzed liability claims related to collisions from 25.3 million fully autonomous miles driven by Waymo in four cities Phoenix, San Francisco, Los Angeles and Austin. They then compared those miles to human driver baselines, which are based on Swiss Re's data from over 500,000 claims and over 200 billion miles traveled. They found that the performance of Waymo's vehicles was safer than that of humans, with an 88% reduction in property damage claims and a 92% reduction in bodily injury claims. Across 25.3 million miles, Waymo was involved in nine property damage claims and two bodily injury claims. The average human driving a similar distance would be expected to have 78 property damage and 26 bodily injury claims, the company says. Waymo's vehicles also performed better when compared to new vehicles equipped with all the latest safety tech, including automatic emergency braking, lane keep assist and blind spot detection. When compared to this group, Waymo's autonomous driving System showed an 86% reduction in property damage claims and a 90% reduction in bodily injury claims. The last time Waymo did a Comparison with Swiss Re's liability data, it only had 3.8 million fully autonomous miles in California and Arizona. Now it has a much bigger data set with 25.3 million miles. The company recently announced that it had driven 4 million trips in 2024 alone. Waymo says it has submitted its latest comparisons with Swiss Re's insurance data to a scientific journal for publication. The data is important because there is still a fierce debate about the safety of driverless cars. Companies like Waymo and others say driverless cars are necessary as an antidote to the crisis of traffic fatalities, of which There are around 40,000 a year in the U.S. they point out that driverless cars never get drunk, tired or distracted and are able to avoid the human errors that so often lead to crashes and deaths. But there is a lack of certainty around the safety of driverless vehicles, mainly because there are far fewer AVs on the road than human driven vehicles, and thus less data from which to draw conclusions. Humans drive close to 100 million miles between fatal crashes, so we're likely to need hundreds of millions of miles from autonomous vehicles before we can start to make more meaningful comparisons about safety. End quote Researchers have unveiled Genesis, an open source generative physics engine that trains robots in simulated reality 430,000 times faster than in the real world, quoting Ars Technica. The accelerated simulation means a neural network for piloting robots can spend the virtual equivalent of decades learning to pick up objects, walk, or manipulate tools during just hours of real computer time. Genesis arrives as robotics researchers hunt for better tools to test and train robots in virtual environments before deploying them in the real world. Fast, accurate simulation helps robots learn complex tasks more quickly while reducing the need for expensive physical testing. The Genesis platform, developed by a group led by Zhou Xian of Carnegie Mellon University, processes physics calculations up to 80 times faster than existing robot simulators like Nvidia's Isaac Gym. It uses graphics cards similar to those that power video games to run up to 100,000 copies of a simulation at once. That's important when it comes to training the neural networks that will control future real world robots. If an AI can control 1,000 robots to perform 1 million skills in 1 billion different simulations, then it may just work in our real world, which is simply another point in the vast space of possible realities, wrote Fan in his X post. This is the fundamental principle behind why simulation works so effectively for robotics. The team also announced the ability to generate what it calls 4D dynamic worlds, perhaps using 4D, because they can simulate a 3D world in motion over time. The system uses Vision language models to generate complete virtual environments from text descriptions similar to prompts in other AI models. Utilizing Genesis's own simulation infrastructure, APIs, to create the worlds the AI generated, worlds reportedly include realistic physics, camera movements and object behaviors, all from text commands. The system then creates physically accurate ray traced videos and data that robots can use for training. This prompt based system lets researchers create complex robot testing environments by typing natural language commands instead of programming them by hand. Traditionally, simulators require a huge amount of manual effort from artists, 3D assets, textures, scene layouts, etc. But every component in the workflow can be automated, wrote Fan. Using its engine, Genesis can also generate character, motion, interactive 3D scenes, facial animation, and more, which may allow for the creation of artistic assets for creative projects, but may also lead to more realistic AI generated games and videos in the future, constructing a simulated world in data instead of operating on the statistical appearance of pixels, as with a video synthesis diffusion model. Genesis remains under active development on GitHub, where the team accepts community contributions okay, so no weekend bonus episodes for you this weekend for the first time ever. Sorry everybody. I'm really limping over the line this week, as you can hear, but two days for my voice to recover and then more because I'll be doing shows Monday and Friday next week with only a bonus episode maybe in between. So by the time I speak to you on Monday, hopefully, but definitely Friday a week from now, I hope to have recovered significantly. Have a good weekend. Talk to you on Monday.