Inside OpenAI's remarkable journey to shape the future of AI

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each weekday, we scour the web to aggregate the many stories related to artificial intelligence; we pass along the news, useful resources, tools or services, technical analysis and exciting developments in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on under the hood in AI.

What OpenAI really wants

From WIRED (may be paywalled): This expansive profile (written by Steven Levy, who has covered the company in depth from their inception) chronicles the remarkable eight-year journey of OpenAI from a struggling nonprofit research lab to a world-changing AI powerhouse. It follows the arcs of visionary characters like Sam Altman, Ilya Sutskever, and the mostly-unknown Alec Radford as they pursue the audacious goal of developing safe artificial general intelligence (AGI).

More details:

• OpenAI floundered initially with unclear direction until Alec Radford's breakthrough on a dataset of Amazon reviews put them on a path to match bigger tech rivals.

• After losing Elon Musk's financial backing, OpenAI made controversial moves like creating a for-profit arm and partnering with Microsoft to gain resources needed to fulfill their mission.

• The release of ChatGPT took OpenAI from niche player to centerstage, forcing society to grapple with implications of powerful AI. OpenAI itself has since transformed into a sort of tech unicorn, with nonprofit roots that still fuel the vision but an army of employees on the for-profit ledger.

• OpenAI makes an intense effort to get ahead of policymakers and frame the public discussion about AI safety. This creates a conflict of interest between development and safety - but Altman has typically charmed key legislators.

Takeaways: It’s hard not to attribute ChatGPT’s surprise emergence mainly to a 10-year-old, nerdy Samuel Harris Altman refusing to let go of a dream to see “a computer that thinks”. His dogged pursuit of that vision (even when many of his peers viewed language models as nothing more than novelties from 2015-2020) led him to the right mix of people for the job, and an unexpected breakthrough provided enough of a spark to light an AI wildfire that has since taken the world by storm.

Looking past the article, OpenAI is currently on a mission (led by Sutskever and Jan Leicke, formerly of Google DeepMind) towards superalignment - the idea that we can keep AI systems that will soon become much more intelligent than us under human control. They are currently devoting 20% of their overall computing power to this project, with the hope that creating a “super researcher” AI will bolster the effort in a way no human(s) could.

If you want to learn more about superalignment, we recommend listening to Leicke himself on the subject. Is their mission crazy, improbable? Jan talks about that. Hopefully his take helps you draw your own conclusions.

Either way, it’s been a wild ride for OpenAI already.

If the chip was either stockpiled or made with smuggled foreign equipment, it suggests the need to tighten export controls. However, if China rapidly innovated the manufacturing process itself, it would indicate the policy has backfired by spurring unwanted tech advancement. A key metric is the manufacturing yield - if low, it's just an expensive demonstration rather than scalable production. Until more evidence appears, the new chip may not represent a major setback to US restrictions intended to thwart China's tech ambitions.

Minnesota-based company Starkey's Genesis AI hearing aids can make what you hear crisper, but they can also serve as a personal assistant. You can use them as a phone, or listen to music, or even to translate languages. There's an accidental fall detection feature - if somebody falls, it'll text up to three people.

• You can now make an AI clone of yourself - or anyone else, living or dead - with Delphi

• Qualcomm focuses on AI and auto as Nvidia takes over as world’s biggest fabless chip company

• Intuit launches generative AI–powered digital assistant for small businesses and consumers

• Oregon Attorney General Rosenblum asks Congress to protect children from artificial intelligence

• (More on yesterday’s Zoom update) Zoom rebrands existing - and intros new - generative AI features

Sustainable magnets are important because they reduce the dependency on materials like cobalt and rare earth elements, which are often linked to environmental degradation and unethical mining practices (in addition to being limited in supply and subject to geopolitical risks). Magnets are used in technologies from wind turbines and electric vehicles to refrigerators and hard drives.

Fusion energy, the process that powers the sun, has been touted as the Holy Grail of clean energy for decades. Unlike fission, which splits atoms and produces long-lasting radioactive waste, fusion combines atoms under high temperature and pressure to create energy with minimal environmental impact.

Making fusion a practical energy source has been a scientific and engineering challenge for decades and previous estimates put the earliest arrivals of fusion energy around 2050 - but 2023 has already been a big breakthrough year for fusion power.

• Hugging Face presents Training Cluster as a service

• (Open source) Open Interpreter allows LLMs to run code locally. Summarize PDFs, visualize datasets, etc.

• VMware’s dev-centered approach to pre-trained models and generative AI

• Introducing LangChain Hub - a place to publish, discover, and try out prompts