Meta Breaks AI Barriers with the Release of Open-Source LLaMa 2

Mornin’ miners⛏️,

Happy Monday!

Welcome to "Digger Insights" - your daily 5-minute enlightenment on the most recent tech updates! In our enjoyable and easily digestible dispatches, we break down the latest tech trends, giving you a quick and comprehensive view.

Join us, and in just a few minutes each day, you'll gain an advantageous perspective on the ever-evolving tech landscape. Ready to dive in?

Let’s get to it!

Today’s Highlights:

• 🖥️Meta Breaks AI Barriers with the Release of Open-Source LLaMa 2🦙

• 💰AI Chip Startup Tenstorrent Raises $100M💸

Meta Redefines the AI Chatbot Innovative Landscape with Open-Source LLaMA 2

The number of companies developing and/or utilizing AI chatbots that are circulating in our tech-heavy society has become endless. Chatbots have stopped becoming mere gimmicks used by companies to show how technologically advanced they are, growing into tools that are extremely useful, practical, and essential for businesses.

Whether used as a method of streamlining customer service processes or automating workloads, it is undeniable that the chatbot market has expanded and will probably continue to grow. According to figures published by Statista, the chatbot market is forecasted to reach $1.25 billion in 2025, increasing significantly from its market size of $190 million in 2016.

Various tech companies have played a part in contributing to this market, and now a recurring player has come for more. After witnessing the success of its large language model, LLaMa (Large Language Model Meta AI), Meta released LLaMA 2, the second version of its versatile and open-source chatbot, only six months after its predecessor.

Open Source and Innovative

Meta CEO Mark Zuckerberg has never been shy of his belief that open-source software drives innovation, rendering it vital in the industry. With Meta’s open-source LLaMA 2 now released, the code and data behind LLaMa 2 can be utilized by researchers worldwide to build upon and improve the technology.

By releasing such software, Meta seems to challenge the restrictive practices done by its other big tech competitors, such as OpenAI with ChatGPT and Google with Bard. Aside from being open-source, the LLaMa 2 is pursued with a different training method. While ChatGPT uses supervised, fine-tuned labeled data provided by human annotators, reinforcement learning* from human feedback (RLHF) is used on LLaMa 2. This technique entails the LLaMA 2 with direct feedback from human testers and users.

Zuckerberg believes that open source can not only motivate more developers to build more advanced technology, but it could actually also improve safety and security for both users and developers. By democratizing the large language model, more individuals can now access, analyze, examine, and inspect the technology, potentially identifying and fixing issues that may or may not occur.

With worldwide access, the Facebook parent company believes that it could rapidly advance the field of AI, as developers now have powerful tools to build and develop innovative applications, each hoped to be beneficial to various different industries.

*Reinforcement learning: a machine learning training method based on rewarding desired behaviors and/or punishing undesired ones, taking suitable action to maximize reward in a particular situation.

Easy, Open Access

The LLaMA 2’s open-source nature has allowed interested parties to interact with it quite easily. The model, which comes in three different sizes of 7 billion, 13 billion, and 70 billion parameters, can be accessed through its demo online, which is hosted by Andreessen Horowitz.

The model can be asked any questions on any topic and be requested creative content using prompts. The demo provides users with three different chat modes, which include balanced, creative, and precise. With these options, individuals can try out the model suited to their preferences. The demo requires an a16z or GitHub login to prevent abuse.

Users who want to run the model on their own machines or modify the code can download LLaMa 2 directly from the AI-sharing platform Hugging Face. Start facing and charging through the world with innovation simply by having a Hugging Face account as well as the required libraries and dependencies to run the code.

Advanced users and programmers who are planning to use the model with more professional and formal intent can also access the LLaMa 2 through Microsoft Azure and Amazon SageMaker JumpStart. Both methods require accounts and subscriptions to the respective services.

One final way of testing out the LLaMa 2 is through a variant of the model through Perplexity.ai, a web crawler that utilizes machine learning to generate answers to users’ queries and provide website links. The Llama.perplexity.ai variant merges the power of the two in just one place.

With so many places to go to try Meta’s groundbreaking open-source chatbot, it wouldn’t be a surprise if more surges of innovative AI applications are seen in the near future. It is an exciting time to be alive and in business, as the LLaMA 2’s immense bounds may unlock great potential for organizations to develop AI applications and solutions tailored to each industry’s needs.

Funding News

Tenstorrent Lands $100M Investment

AI chip startup Tenstorrent has just raised $100 million in a convertible note funding round. This Toronto-based startup sells AI processors for deep learning and licenses AI software solutions. Aside from this, Tenstorrent also licenses IP around RISC-V, the open-source instruction set architecture used to develop custom processors.

The startup, founded in 2016 by Ivan Hamer, Ljubisa Bajic, and Milos Trajkovic, all former employees of AMD, started out as a company that developed its own in-house infrastructure. Tenstorrent announced an all-in-one system used to accelerate AI model training in data centers and public and private clouds named Grayskull. However, the company shifted its focus and launched DevCloud, a cloud-based service for developers to run AI models without hardware, in 2021. Though its imagined all-in-one system hadn’t come true, the company now presents its products with its name, in a way still respecting its memory and legacy.

Tenstorrent, now managed by engineer Jim Keller, raised its funds with the support of Hyundai Motor Group and Samsung Catalyst Fund.

Hyundai’s two car-making units, Hyundai Motor and Kia, took part with $30 million and $20 million, respectively, to develop CPUs and AI co-processors for future mobility vehicles and robots. Samsung Catalyst Fund, along with other VCs such as Fidelity Ventures, Eclipse Ventures, Epiq Capital, and Maverick Capital, contributed a total of $50 million.

Tenstorrent plans to allocate a total of $334.5 million already raised from all its funding rounds so far to product development, the design and development of AI chiplets, and its machine learning software roadmap.

Meme & AI-Generated Picture

Job Posting

• MassMutual - Security Architect - Boston, MA (Hybrid)

• Gopuff - Facilities Administrator - Philadelphia, PA (Remote)

• Litera - Lead UX Researcher - United States (Remote)

• Elementl - Product Designer - Minneapolis-Saint Paul, MN (Remote)

Promote your product/service to Digger Insights’ Community

Advertise with Digger Insights. Digger Insights’ Miners are professionals and business owners with diverse Industry backgrounds who are looking for interesting and helpful tools, products, services, jobs, events, apps, and books. Email us [email protected]

Your feedback would be greatly appreciated, send it to [email protected]