Ashley Stewart Business Insider

Microsoft unifies Copilot under one team and moves Mustafa Suleyman to focus on superintelligence.

by with No Comment Xpandfy

Microsoft is continuing a major leadership shakeup by creating a new combined team for its commercial and consumer Copilot products, under a new executive. The company is also moving Microsoft AI CEO Mustafa Suleyman to its superintelligence team, focused on building frontier AI models.

The reorganization underscores how high the stakes have become to dominate AI in Big Tech. Microsoft is trying to solve two problems at once: turning its sprawling lineup of Copilot products into a coherent platform for both businesses and consumers, while also reducing its reliance on OpenAI by building its own frontier models.

The structural changes also reflect Microsoft CEO Satya Nadella’s ongoing strategy to reorganize the company to compete more aggressively in the AI race.

Microsoft’s many Copilot products have been a source of confusion for customers, and the company has been trying to address that by creating a cohesive experience across applications, according to a recent all-employee town hall viewed by Business Insider.

“This is how we move from a collection of great products to a truly integrated system, one that is simpler and more powerful for customers,” Nadella wrote in an internal email announcing the reorganization.

Jacob Andreou, whose previous title was Microsoft AI corporate vice president of product and growth, will become executive vice president of Copilot and lead the combined teams.

Microsoft recently promoted executives Ryan Roslansky, Perry Clarke, and Charles Lamanna to take over for longtime executive Rajesh Jha, who earlier this month announced his retirement. Those executives will lead Microsoft 365 apps and the Copilot platform, and make up the Copilot leadership team with Suleyman and Andreou.

Suleyman in an email to employees announcing the changes said he will now focus all of his “energy on our Superintelligence efforts and be able to deliver world class models for Microsoft over the next five years.”

Suleyman formed a superintelligence team at Microsoft in November focused on training “frontier models of all scales with our own data and compute at the state-of-the-art level” to make the company “self-sufficient in AI,” he told Business Insider at the time.

Microsoft’s previous deal with OpenAI barred the software giant from developing its own AGI through 2030, according to a person familiar with the matter. A new deal announced in October allowed the companies to “independently pursue AGI (artificial general intelligence) alone or in partnership with third parties.”

Read the memos

Nadella’s memo:

I want to share two org changes we’re making to our Copilot org and superintelligence effort.

It’s clear a new era of productivity is emerging as AI experiences rapidly evolve from answering questions and suggesting code, to executing multi-step tasks with clear user control points. You see this in our announcements over the last couple of weeks, like Copilot Tasks and Copilot Cowork, agentic capabilities in Office, and Agent 365. As these experiences connect more naturally across agents, apps, and workflows, we have an opportunity to help customers spend more time on higher-value work and reduce manual coordination, while providing people with more agency and empowerment and organizations with the governance and security controls they need.

To that end, we are bringing the Copilot system across commercial and consumer together as one unified effort. This will span four connected pillars: Copilot experience, Copilot platform, Microsoft 365 apps, and AI models. This is how we move from a collection of great products to a truly integrated system, one that is simpler and more powerful for customers.

Jacob Andreou will lead the Copilot experience across consumer and commercial, driving design, product, growth, and engineering, as EVP, Copilot, reporting to me. As CVP of Product and Growth at Microsoft AI, Jacob has accelerated our user-focused AI-first product making and growth framework. Prior to that, he was SVP at Snap, where he helped scale the company from its early days.

Progress at the AI model layer is more critical than ever to our success as a company over the next decade and is foundational to everything we build above it. We are doubling down on our superintelligence mission with the talent and compute to build models that have real product impact, in terms of evals, COGS reduction, as well as advancing the frontier when it comes to meeting enterprise needs and achieving the next set of research breakthroughs. Mustafa Suleyman and I have been working towards this plan for some time, and he will continue to lead this high ambition work, reporting to me. Mustafa is uniquely qualified to drive this forward, with his deep focus and commitment to advancing the frontiers of model science, while also ensuring that human control, agency, and economic opportunity remain at the center of these advancements.

Ryan Roslansky, Perry Clarke, and Charles Lamanna will lead M365 apps and the Copilot platform. Together, Jacob, Ryan, Charles, Perry, and Mustafa make up the Copilot LT and over the next few weeks they’ll work to align the teams.

Our org boundaries will simply reflect system architecture and product shape such that we can deliver more coherent and competitive experiences that continue to evolve with model capabilities. And I am looking forward to how together we apply all of this to empower people, organizations, and the world.

Suleyman’s memo

Technology and the future of our industry will be defined by two things: frontier models, and the products through which they are experienced. For some time, I’ve been thinking about how we best tackle these huge challenges, and today I’m excited to be evolving our structure at MAI, ensuring we’re positioned to succeed in both.

I came to Microsoft with an overriding mission: to create Superintelligence that delivers a transformative, positive impact for millions of people. This requires us to build frontier models, at scale, pushing the boundaries of what’s possible. Everything else follows from this. It’s the foundation for our future as a company. With our ambitious, long-term frontier scale compute roadmap locked, we now have everything we need to build truly SOTA models.

As you will have just heard from Satya, the next phase of this plan is to restructure our organization to enable me to focus all my energy on our Superintelligence efforts and be able to deliver world class models for Microsoft over the next 5 years. These models will enable us to build enterprise tuned lineages that help improve all our products across the company. They’ll also enable us to deliver the COGS efficiencies necessary to be able to serve AI workloads at the immense scale required in the coming years. Achieving all this will be a huge challenge, and I’m committing everything we have — and I have personally — to make it happen.

To that end, I’ve been working hard with other leaders in the background for a while now to define a strategy to unify Copilot by bringing together the Consumer and Commercial efforts as one. We all know this makes sense. Every user — whether at home or at work — will be able to enjoy the full benefit of what we are all building. Today, we’re combining these organizations into a single, unified Copilot org. Jacob has demonstrated himself to be an outstanding leader for the product experience and clearly has the product instincts, the operational range, and the conviction to make Copilot a great success.

Jacob will retain a dotted line to me, and I’ll stay directly involved in much of the day-to-day operation of MAI, attending Meetups, MMMs, LT, and supporting Jacob to drive all areas of product strategy. To ensure that the models we build and the products we ship are mutually reinforcing, we are establishing a Copilot Leadership Team that includes me, Jacob, Charles Lamanna, Perry Clarke, and Ryan Roslansky. This will enable us to focus our brand strategy, our product roadmap, our models and our core infrastructure as one to deliver the best experiences possible for all our users.

Thank you for everything you’ve done over the last few years. I know how hard everyone has been pushing and the sacrifices many of you have made to help the company adapt to this new era.

We really do have an incredible opportunity to redefine Microsoft for this agentic revolution

Mustafa




Source link

Chong Ming Lee, Junior News Reporter at Business Insider's Singapore bureau.

I work at Meta’s Superintelligence Labs and used to be at OpenAI. Here’s what the job is like — and what I’ve learned.

by with No Comment Xpandfy

This as-told-to essay is based on a conversation with Prakhar Agarwal, an applied researcher at Meta Superintelligence Labs who previously worked at OpenAI. The following has been edited for length and clarity. Business Insider has verified his employment and academic history.

My day-to-day varies a lot depending on what stage of the project we are in versus what the immediate deliverables are.

At OpenAI and Meta, you have these milestones — say, a big training or reinforcement-learning run — in 10 months. It gets intense when we’re approaching the deadline.

Whatever work I identify is always based on the current iteration of the model. If I say the model isn’t good at X and my solution helps fix X, it is based on that version of the model. If I miss the deadline, I don’t know whether the next version will have the same issues or not.

If we are further away from that deadline, then we’re mostly working on evaluations and trying to find failure cases and issues with the existing model.

The work is super dynamic. Sometimes you think something is super easy and you’ll get it done in a day. Other times, it’s the opposite — because there are so many unknowns, it might take a week.

Working at frontier labs feels very different from Big Tech

What we’re limited by in these foundational labs is compute. It’s not like Big Tech or other places where you can keep hiring a bunch of people and give them small pieces of a task to do.

Everyone needs compute to actually do something, and as soon as you have a lot of people, the compute gets divided, so no one will be able to do anything.

You also want high-bandwidth communication between stakeholders — you don’t want 10 different layers of communication. The speed of iteration is much faster. These core groups tend to be much smaller and tighter.

The idea of a “team” is also very fluid. Each person has their own projects, but they collaborate with others to work on joint projects. At Meta and OpenAI, there are a lot of senior people and not a lot of junior people, so everyone has a decent scope of projects.

Sometimes I collaborate more with people outside my immediate team than within it. Your scope isn’t restricted to four or five people. Your scope is the problem you’re trying to solve.

Communication and going deep with coding are key

Communication is the most important aspect in these labs. Because a lot of things aren’t documented, you need to be able to articulate what you’re doing, why you’re doing it, what the next steps are, convey your results, and get feedback on your work.

Becoming comfortable going through the code and identifying the specifics is one of the most important skills I’ve seen. The speed at which the code evolves is much faster than the documentation. If you’re stuck on something, read the code and try to understand it yourself.

Having some understanding of what’s happening across different verticals also gives you a good overview of the ideas and approaches people are trying. Because everything is super related, you might learn something from there or find ways to contribute.

The biggest advantage these labs have is knowing what doesn’t work

A research paper tells you, “I did X, Y, and Z in this specific order, and it works.” But what you don’t see is that before doing X, Y, and Z, I tried 50 different things that didn’t work — and people don’t talk about that.

That, to me, is the real strength of these foundation labs. Because of all the experimentation and all the work that has already been done, the teams have built really strong intuitions. They know which things won’t work or won’t scale, and which are going to work well.

People outside often look for the gains, but they miss the point that even the misses are very valuable.

Advice for those who want to work in top labs

I don’t have a good answer for managing burnout. You’re pretty much just going with the flow. You’re working at the cutting edge, and to put it simply, if you want to be here, you can’t think about it on a strict day-to-day basis.

What I would tell my younger self is to be comfortable exploring new avenues and new ideas. What I’ve seen is that we try to play to our strengths or stay in a deterministic setting where we know we’ll do fine. But in these domains, the speed at which things are moving is so fast that you need to be able to switch to a new topic.

Build the muscle to handle being thrown into something completely new. Sometimes, it’s more psychological than a skill issue.

Do you have a story to share about working at a top AI lab? Contact this reporter at cmlee@businessinsider.com.




Source link

Pranav Dixit

Meta is forming a new AI engineering org for its superintelligence push, with teams as large as 50 people per manager

by with No Comment Xpandfy

Meta is establishing a new applied AI engineering organization designed to accelerate the company’s push toward superintelligence, according to two employees familiar with the matter.

The new organization will be headed by Maher Saba, a vice president at Reality Labs, the division responsible for Meta’s metaverse products and AI-powered smart glasses. Saba’s new group will report directly to Chief Technology Officer Andrew Bosworth. Teams within the organization will have manager-to-employee ratios of up to 1:50, the people said.

Meta declined to comment.

The group will work in close partnership with Meta Superintelligence Labs, the organization that Meta created last summer and is led by former Scale AI chief Alexandr Wang, to oversee the development of Meta’s frontier AI models. Saba’s team will build “the data engine that helps our models get better, faster,” according to an internal memo, sources said. The Wall Street Journal first reported about the memo.

The new organization will have two distinct teams: one focused on building interfaces and internal tooling, and another dedicated to helping feed the AI with data.

Saba wrote in the memo that “building great models isn’t just about researchers and compute,” according to the employees.

Saba added that the group aims to turn capable AI models into market-leading ones. He pointed to recent AI research gains in reinforcement learning and post-training as evidence that Meta has an opening to accelerate if it invests more aggressively in this area, the people said.

The unusually flat structure reflects a broader organizational philosophy that CEO Mark Zuckerberg outlined during Meta’s most recent earnings call. Zuckerberg told investors that Meta is “elevating individual contributors and flattening teams” and said the company is already seeing “projects that used to require big teams now be accomplished by a single, very talented person.”

Another Big Tech company, Nvidia, is also known for its flat structure, with CEO Jensen Huang having over 30 direct reports.

Have a tip? Contact Pranav Dixit via email at pranavdixit@protonmail.com or Signal at 1-408-905-9124. Use a personal email address and a nonwork device; here’s our guide to sharing information securely.




Source link

Image of Lakshmi Varanasi

This is the key breakthrough AI still requires to reach superintelligence, according to those building it

by with No Comment Xpandfy

In humans, working memory — our ability to hold and use information in everyday life — is closely linked to general intelligence.

That means the ability for AI to remember things could be the key to realizing a superintelligent AI, a still theoretical version of AI that reasons as well or better than humans.

OpenAI CEO Sam Altman thinks it’s hard to predict just how intelligent AI can really be because the possibilities of memory retention are limitless.

“Even if you have the world’s best personal assistant, they don’t, they can’t remember every word you’ve ever said in your life, they can’t have read every email, they can’t have read every document you’ve ever written, they can’t be looking at all your work every day and remembering every little detail, they can’t be a participant in your life to that degree. No human has like infinite, perfect memory,” Altman said recently on the “Big Technology” podcast.

AI, however, will definitely have the capacity for that, he said.

“Right now, memory is still very crude, very early,” he said. Once AI is able to remember every granular detail of a user’s life, including even the small preferences they didn’t explicitly indicate, it will be “super powerful,” he said.

Altman added that it’s one of the future features he’s most excited about — and he’s not the only one.

Andrew Pignanelli, the cofounder of The General Intelligence Company of New York, a company that builds AI agents for businesses, said that memory will become the biggest focus for AI companies in the coming year.

“It will become the most important topic discussed and recognized as the final step before AGI,” Pignanelli wrote in a blog post. “Every model provider will add and improve on memory for their apps after seeing OpenAI’s success with ChatGPT memory (like Claude just did).”

Pignanelli, however, said that the industry is still a long way from perfecting long-term memory.

“Larger context windows continue to improve things, as they allow more data to be passed into the context window, which allows the agent to better read parts of a large memory index,” he wrote, in reference to the amount of information a large language model can process in a single prompt. “Even then, though, the vast level of detail that we need to reach to consider something AGI requires memory architecture improvements.”

Even shorter-term episodic memory hasn’t been fully solved yet, he said.

Solving that memory problem is the ticket to turning AI from something that feels artificial to something that seems human, he said.

“Our systems today get the interaction part right. In terms of a Turing test for interaction, we’re basically all the way there. But that’s only half of what’s needed to make a digital self,” he wrote.

“The first AGI will be a very intelligent processor combined with a very good memory system,” he said.




Source link