OpenAI's plans according to Sam Altman (4 minute read)
OpenAI is currently extremely GPU-limited and this is delaying a lot of its short-term plans. In the near term, the company plans to make GPT-4 cheaper and faster, create longer context windows, finetune its API, and make a stateful API that remembers conversation history. OpenAI plans to introduce GPT-4 multimodality in 2024. Plugins are probably not coming to the API anytime soon as their usage suggests that they don't have product-market fit yet. OpenAI's internal data suggests that making models larger will continue to yield performance.
|
Apple preps first Mac with M2 Ultra chip and M2 Max desktop ahead of WWDC (3 minute read)
Apple is expected to announce several new Macs at the WWDC 2023 keynote on Monday. The company is testing two Mac desktops with the M2 Max and M2 Ultra chips ahead of the event. The M2 Ultra will have a 24- or 76-core CPU, 60-core GPU, and up to 192GB of RAM. Apple will begin accepting the Mac Studio and M2 MacBook Air into its trade-in program starting next week. The first M3 Macs are not expected to be ready until the end of the year.
|
Science & Futuristic Technology
|
How JetZero aims to be the SpaceX of aviation (13 minute read)
JetZero's blended wing body aircraft could be the first major redesign of a commercial passenger jetliner to enter production. Its lighter weight and superior aerodynamics deliver the same speed and range as existing midbody jetliners on half the fuel. The design could help the airline industry hit emission goals while lowering operational costs. JetZero is aiming for an early-2030s entry to service as commercial jetliners for airlines. Renders of JetZero's wing body aircraft are available in the article.
|
Longevity enthusiasts want to create their own independent state (8 minute read)
A group of longevity enthusiasts wants to create their own independent state to study treatments to slow down or reverse aging. They believe that companies should not be held back by laws that limit how they develop and test drugs. Lowering the requirements for testing longevity treatments in humans could accelerate the development of the technology. The group is considering creating a 'longevity state' in Rhode Island or somewhere in Latin America. There are no firm plans for the state yet as the organizers want any decisions to be made collaboratively.
|
Programming, Design & Data Science
|
Chainlit (GitHub Repo)
Chainlit allows users to create ChatGPT-like UIs on top of any Python code. It features intermediary steps visualization, element management and display, cloud deployment, and more. Plug-and-play integration with LangChain is available.
|
Feedback: I try to answer "how to become a systems engineer" (4 minute read)
Systems engineers design, integrate, and manage complex systems over their life cycles. Being curious about things and how they work is a helpful mindset for a systems engineer. Trying to work out why things are the way they are will result in working models of why things are the way they are. The experience gained when working out these systems will help when dealing with other systems issues in the future.
|
Would Large Language Models Be Better If They Weren’t So Large? (7 minute read)
Large language models that generate text improve as they are fed more data. Making bigger and more capable AI requires processing power that few companies possess. There is a growing concern that a small group of tech companies will exercise near-total control over the technology. A group of academics working in natural language processing has issued the BabyLM Challenge to turn this paradigm on its head. The group is calling for teams to create functional language models using data sets that are less than one-ten-thousandth the size of those used by the most advanced language models.
|
OpenAI is pursuing a new way to fight A.I. ‘hallucinations’ (3 minute read)
OpenAI has developed a new method for training artificial intelligence models that hallucinate less. AI hallucinations occur when large language models fabricate information while behaving as if they are spouting facts. Even state-of-the-art models are prone to hallucinations. The new training method awards models for each individual correct step of reasoning instead of only rewarding correct final answers. While OpenAI did not invent the approach, it is helping to push it forward. |
localGPT (GitHub Repo)
localGPT uses the Vicuna-7B model and InstructorEmbeddings to allow users to ask questions about their documents without using an internet connection.
|
TLDR Talent is our exclusive community where we help world-class tech talent and get intros to companies of their choice, along with a number of exciting startups and tech companies curated by TLDR.
We give you full control of the process, you can specify if you’re actively searching or passively interested only if something amazing comes along. Set your preferred compensation, seniority/title/role, specific companies (or types of companies) you’d like to work for and more. Click here to apply.
|
If your company is interested in reaching an audience of tech executives, decision-makers and engineers, you may want to advertise with us.
If you have any comments or feedback, just respond to this email!
Thanks for reading,
Dan Ni and Stephen Flanders
|
|
|
|