Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Google, Cloudflare, and Cisco have been ordered to implement measures to prevent French internet users from using their services to access around 117 pirate domains.
OpenAI's newest board member Paul M. Nakasone, a retired US Army general and the former director of the National Security Agency, will also join OpenAI's recently created Safety and Security Committee.
Granting foundation models the ability to think for longer might upend Scaling Laws and change AI's trajectory.
The disorganized sharp-wave ripples of sleep deprived rats seem to prevent them from effectively flagging experiences for long-term memory.
Sponsored
WorkOS helps companies quickly become ENTERPRISE READY.
This article provides a summary of what was discussed at Tesla's annual meeting. Tesla shareholders have voted to ratify CEO Elon Musk's $56 billion pay package. While the vote doesn't override an earlier court ruling that ordered Tesla to rescind the package, it provides a public relations victory for Musk and could help him sway a court to give him the performance options in the future. Shareholders also voted to move Tesla's site of incorporation out of Delaware and into Texas.
Amazon demoed a generative AI-powered Alexa in September last year. There has been little news about the project since, besides some reports about a potential launch later this year. Shipment of the new Alexa is being delayed due to structure dysfunction and technology challenges within Amazon. Former Amazon employees say that the company is desperately behind its Big Tech rivals in the race to launch AI chatbots and agents. The new Alexa is reportedly still not ready for a prime time rollout and Amazon does not have enough data or access to the specialized chips needed to run large language models to compete. Amazon so far has been unable to capitalize on its partnership with Anthropic to build a better Alexa.
A group of researchers have collaborated with scientists at Google's DeepMind AI lab to build a biomechanically realistic digital model of a rat. They trained an artificial neural network to control a virtual body in a physics simulator using high-resolution data recorded from real rats. This resulted in a virtual rat with an artificial brain that can move around virtual environments like a real rodent. The research could provide convenient and fully transparent models for studying neural circuits and see how such circuits are compromised in disease.
Computers can help solve the big, unsolved problems in mathematics. The latest advancements in artificial intelligence introduce completely new ways of working and open up new possibilities for cooperation in mathematics. This article contains a transcript of an interview with Terence Tao from the University of California in which he discusses the future of the field of mathematics. Tao also covers collaboration in the field, automated proof checking, new technologies, the big unanswered problems in mathematics, and more.
Sponsored
Everyone wants to scale faster, but few know how. Register for Effects, Plaid's annual virtual conference [9], to hear how today's top executives are growing their businesses–and be the first to hear about the new tech shaping the future of finance. Don't miss this free event happening this Tuesday. Register today [9].
Smart Paste is a tool broadly available at Google that predicts the next state of a code environment and uses generative AI to create context-aware adjustments to pasted code. It streamlines the process of copy-pasting during code revisions. When testing the tool with around 40,000 engineers, Google found that 6.9% of all pastes in the IDE utilized Smart Paste - 42.5% of suggestions from the feature were accepted. This article discusses how Google developed the feature.
Abliteration is a technique that uncensors any large language model without retraining. It effectively removes the model's built-in refusal mechanism, allowing it to respond to all types of prompts. Large language models that are fine-tuned for safety and instruction-following have a stream that moderates the model's ability to refuse requests. Preventing models from representing this direction makes the model lose its ability to refuse requests. Adding to the direction artificially can cause models to refuse even harmless requests. This article discusses the abliteration technique and walks readers through how to implement it.
The jhanas are a series of altered mental states induced via sustained concentration without any external stimuli or substances. These states vary in intensity, with some comparable to drug-induced states. Practicing the jhanas can be a good way to cultivate attention. This guide teaches readers how to enter these states. Anyone can access the jhanas regardless of their meditation experience.
Dream Machine is a next generation video model from Luma AI. It can create high quality, realistic shots from text instructions and images. Dream Machine is available to everyone for free. Examples of videos generated with Dream Machine along with a link to try the model are available.
Sponsored
Instantly set up a two-way integration between Jira Service Management and your company chat apps. Keep users happy and Jira
X is planning to implement a broader and more robust promotion process that will involve lighter-weight check-ins throughout the year.
Dave Burke, the VP of engineering for Android, is shifting into an advisory position while exploring AI/bio projects.
Instead of pushing people to be typing all the time, managers should instead provide the materials, environment, and processes necessary to ensure that the thinking developers do is of high quality.
Interaction to Next Paint became part of Google's Core Web Vitals in March.
bignums are very large numbers that allow developers to go beyond CPU limitations for representing integers and performing arithmetic.
shpool is a service that enables session persistence.
Sponsored
IPOs are serious business — and you'll need your financial operations to be efficient, error-free, and auditable well in advance.
Apple and OpenAI's partnership isn't expected to generate meaningful revenue for either party, at least at the beginning. Apple believes that pushing OpenAI's brand and technology to its devices is of equal or greater value than monetary payments so it isn't paying OpenAI as part of the deal. The deal isn't exclusive - Apple is already discussing offering other chatbots as additional options. Apple eventually plans to make money from AI by striking revenue-sharing agreements with AI partners.
Meta is reportedly looking to cut down on its number of vice-president positions from 300 to about 250. The company's job cuts were guided by principles focused on making Meta an even stronger technology company. These principles included removing multiple layers of management, canceling projects, making every organization leaner, building an optimal ratio of engineers to other roles, and investing in AI and other tools. Meta's capital expenditures on AI and Reality Labs will range between $35 billion and $40 billion by the end of 2024.
A recent tweet from Tesla's official account listed a rundown of the company's accomplishments under Elon Musk's leadership since 2018, when the $55 billion pay package was approved by shareholders. The last item on the list claims that the company has two Optimus robots performing tasks in a factory autonomously. Musk recently said that the Optimus robot might go on sale as soon as next year. It is expected to cost 'less than half of a car'.
Cerebras, a California-based company, has demonstrated that its second-generation wafer-scale engine is significantly faster than the world's faster supercomputer in molecular dynamics calculations. It can also perform sparse large language model inference at one-third of the energy cost of a full model without losing any accuracy. Both achievements are possible due to the interconnects and fast memory access enabled by Cerebras' hardware. Cerebras is looking to extend the applications of its wafer-scale engine to a larger class of problems, including molecular dynamics simulations of biological processes and simulations of airflow around vehicles.
Sponsored
Data brokers build shadow profiles of you that include a frightening amount of personal details. They sell them to scammers, insurance companies, banks, businesses, and even the government. And to stop this, you'd need to contact hundreds of data brokers… or just use Incogni [10]. 3 minutes to set up, 180+ data brokers tackled. Use code
Restate is a platform for building resilient applications that tolerate all infrastructure faults. It enables developers to easily build workflows, event-driven applications, and distribute services in a fault-tolerant manner with durable async/await. Restate is great at building lambda workflows as code, transactional RPC handlers, event processing with Kafka, and more. It supports TypeScript, Java, and Kotlin.
NLUX is a conversational AI JavaScript library that provides a UI for large language models. It makes it super simple to integrate powerful large language models into web apps. NLUX features React components and hooks, LLM adapters, streaming LLM output, and custom renderers.
Founder liquidity is the practice where founders sell a portion of their shares during a funding round. It allows founders to reduce risk and secure personal financial stability while continuing to build the company with a fresh influx of venture capital. The practice is often kept secret because it undermines the narrative of the founder who is 'all-in'. It might change how startups are perceived and valued if it were widely known that founders could de-risk their financial position while their employees remained all-in.
Competitive Excel is a real thing. The Excel World Championship was recently held at the MGM in Las Vegas. This article takes readers to the front rows of the event. It details what the competition entails and what the event was like. While the event may never be mainstream, it proves that spreadsheeting can be a sport.
Sponsored
For the next 12 hours, try Dollar Flight Club for just $1. [15] Fly to Hawaii from $109. Greece from $247. Or Mexico from $137. The $1 deal is only here for 12 hours. Join 1 million members [15] for just $1 today.
10,900 video game industry employees have been laid off so far in 2024 compared to the prior record of 10,500 in 2023.
Systems powered by Nvidia's Hopper architecture dominated the results of two new tests from MLPerf, an AI benchmarking suite, that compare the fine-tuning of large language models and training of graph neural networks.
Elon Musk responded to Apple's integration of OpenAI ChatGPT on some of its devices by threatening to ban all Apple devices from the premises of his companies.
The vast majority of the revenue comes from OpenAI's products and services - the company has been working to boost revenue by selling services to enterprise customers.
Firefox will start automatically upgrading audio, video, and image subresources from HTTP to HTTPS starting with version 127.
This article contains a transcript of an interview with Aidan Gomez, CEO of Cohere, where he discusses Cohere's path to profitability, the benefit of competition in the enterprise space, AI's current abilities, the future of large language models, and more.
designed to support active hydration and a healthy lifestyle.
Sponsored
How do you optimize your performance, physically and mentally? Start by realizing that most people need MORE salt and LESS water.
Many of OpenAI's early employees are sitting on millions of dollars worth of equity. The only way for them to realize any value from this equity in the near term is through secondary stock sales. OpenAI has told employees that it will try to hold one tender offer roughly every year depending on how both the company and the market are faring at the time. Current and former employees are increasingly concerned about access to liquidity after reports that the company has the power to claw back vested equity. The company has used aggressive tactics to get employees to sign exit agreements that affect the future of their stock holdings.
X is rolling out private likes - what users like on the platform will be hidden by default. Owner Elon Musk said that it was important to allow people to like posts without getting attacked for doing so. Users will still be able to see who liked their posts and the like count for all posts.
The experimental drug donanemab slowed the progression of Alzheimer's disease by about 35% in a study of more than 1,700 people. Eli Lilly, the company that makes the drug, has submitted the results of its study to the Food and Drug Administration, which is expected to make a decision on the drug by the end of the year. The drug is still not a cure - its benefits amount to only about a seven-month delay in the loss of memory and thinking. The drug caused dangerous swelling or bleeding in the brain in about 25% of patients.
Stoke Space has fired its first stage rocket engine for the first time, briefly igniting it for about two seconds. The test was a success as the engine performed nominally. Stoke's two-stage Nova rocket will be fully reusable and have a lift capacity of about 5 metric tons. It could be ready for its first flight in 2026.
Sponsored
Looking to hone your craft, advance your DevRel career, or connect with the wider dev tool community? DevRelCon [10] is the place for you. Sessions will cover practical topics [10] like metrics and ROI, managing up and down, community management, content creation, career paths, open source, improving and scaling using AI tools, and much more.
Software is an apprenticeship industry - you can only learn by doing. The industry changes fast, so developers have to keep learning. It takes many years to forge a competent software engineer. Senior software engineers don't only write code - they have to be able to understand, maintain, explain, and manage a large body of software in production over time and translate business needs into technical implementation. Writing code is the easiest part of software engineering, and it's getting easier every day. AI can help generate lots of code really fast, but it can't aid in the work of managing, understanding, or operating that code. Generated code can't be trusted - it needs to be reviewed before it can be committed and shipped to production.