What Is Adept AI Labs?
A Human-Centric Path to AGI
Image: Adept Team, Twitter.
It’s time for a startup story. Greylock Partners is one of the oldest venture capital firms, founded in 1965, with committed capital of over $3.5 billion under management. The firm focuses on early-stage companies in the consumer, enterprise software and infrastructure sectors as well as the semiconductor sectors.
Reid Hofman, one of the co-founders of LinkedIn, recently said this:
As artificial intelligence continues to amplify human capabilities, key industries, and our shared future, we at Greylock are delighted to work with Adept AI Labs.
Who Are Adept Labs?
Adept is building a natural language interface to everyday software tools.
On April 26th, 2022 Greylock was excited to share their investment in Adept, which is coming out of stealth and announcing a $65M Series A co-led by Greylock and Addition. Adept is a ML research and product lab building general intelligence by enabling humans and computers to work together creatively.
AI Transformer Inventors Launch Adept with $65M to Lend a Hand to Knowledge Workers
Adept’s co-founders, CEO David Luan, CTO Niki Parmar and chief scientist Ashish Vaswani, boil their ambition down to perfecting an “overlay” within computers that works using the same tools people do. This overlay will be able to respond to commands like “generate a monthly compliance report” or “draw stairs between these two points in this blueprint,” Adept asserts, all using existing software like Airtable, Photoshop, Tableau and Twilio to get the job done.
Where Will Transformers Lead?
Over the past several decades, and in the last several years in particular, we’ve seen significant advancements in AI and its impact on the way in which we work and live. While these advancements have historically been incremental and task-specific, the invention and introduction of the Transformer in 2017 at Google Brain was a breakthrough for the field – and dramatically accelerated the pace of progress in AI. Transformers excel at generality and can extend across domains – finally making the north star of artificial general intelligence (AGI) tangible.
Who Else Are in on the Round?
The round includes participation from Root Ventures and angels including Scott Belsky (founder of Behance), Howie Liu (founder of Airtable), Chris Re (Stanford), Andrej Karpathy (head of Tesla Autopilot), and Sarah Meyohas.
So immediately there’s a lot of credibility here.
At a product lab called Adept that emerged from stealth in late April with $65 million in funding, they are — in the founders’ words — “build[ing] general intelligence that enables humans and computers to work together creatively to solve problems.”
So it’s interesting to hear the AGI connection and is eerily similar to Reid’s own new Startup, Inflection, that will aim to develop AI software products that make it easier for humans to communicate with computers.
Niki Parmar previously worked in the Google Brain team where she researched on novel techniques in deep learning.
“[We’re training a neural network to use every software tool in the world, building on the vast amount of existing capabilities that people have already created.”
Ashish Vaswani was also at Google brain and started work on Adept around December, 2021.
With UiPath’s stock plummeting in recent times it’s interesting to see something around Voice-AI and RPA. What Adept is creating sounds a little like robotic process automation (RPA), or software robots that leverage a combination of automation, computer vision and machine learning to automate repetitive tasks like filing forms and responding to emails. But the team insists that their technology is far more sophisticated than what RPA vendors like Automation Anywhere and UiPath offer today.
A Google Brain Spin-Off
Since the introduction of the Transformer architecture, we’ve seen incredible models get built that demonstrate the wonderful power of this architecture with high scales of data and compute.
Greylock goes on to state: As just a few examples – GPT-3 can generate text to write poetry or answer emails, DALL-E can generate highly realistic images from natural language descriptions, Codex can assist in the development of high quality software code, and BERT is now a foundational part of powering Google Search.
Yet, despite these dramatic advances, these models have a significant limitation – at least so far.
While excellent at understanding and generation, these models can’t act in the increasingly digital world in which they operate. This startup and team believes the next quantum leap in AI advancement and impact will combine understanding with actuation to transform the way we live and work. Greylock is thrilled to back Adept in that ambition.
What Is Adept?
Adept is an ML research and product lab building general intelligence by enabling people and computers to work together creatively.
They believe that AI systems should be built with users at the center — their vision is one where machines work together with people in the driver's seat: discovering new solutions, enabling more informed decisions, and giving us more time for the work we love.
Hoffman sold LinkedIn to Microsoft for $26.2 billion in 2016 so he knows a thing or two about growing companies and how venture capital works today.
I like how Reid Hoffman phrased this:
Adept takes a different path from other AGI companies. Rather than building general intelligence to take over valuable tasks, they're building AI tools that empower humans to get stuff done.
I mean intuitively that makes sense though not sure about using the term AGI. Don’t ask me whose dog that is.
Adept isn’t the only one exploring this idea. In a February paper, scientists at Alphabet-backed DeepMind describe what they call a “data-driven” approach for teaching AI to control computers. VoiceAI is going to get a lot better at help people in how they work.
In the PR of startups like these, you get glimpses also of the evolution we have achieved in A.I. in recent years.
Adept on their blog say that machine learning has seen more progress in the last five years than in the prior 60. Since the beginning, they’ve wanted to build models with similar plasticity to human intelligence – models that can learn and grow in capability across a highly diverse set of tasks. For most of this time, their best results were limited to models that were engineered to excel in specific domains — they showed promising levels of capability, but were bespoke.
Of course I think they are talking about human-centric AI, nothing like real AGI. It’s important to point this out as companies like OpenAI and their employees have been playing on words about AGI.
Basically by having an AI observe keyboard and mouse commands from people completing “instruction-following” computer tasks, like booking a flight, the scientists were able to show the system how to perform over a hundred tasks with “human-level” accuracy.
David Luan goes on to say that: the Transformer was the first neural network that seemed to “just work” for every major AI use case – “it was the research result that convinced me that general intelligence was possible.” Transformers quickly became the fundamental architecture of giant models with highly general capabilities, giving researchers the key to unlock decades-old problems in rapid succession.
Let’s not forget that David was VP of Engineering at OpenAI for a good two years not too long ago. He’s interesting in that while his main focus is research, most of his career has revolved around near- and long-term impacts of AI on society. Not surprising then that he is doing his own thing.
You might remember a while back when Siri unveiled Viv, an AI platform that promised to connect various third-party applications to perform just about any task. That didn’t quite work out but maybe Adept can do something similar to that promise.
I sort of get why David might see how this could lead to AGI parallels.
David will evolve to be pretty important to the future of A.I. policy and regulation. Reid Hoffman insists that “Developing AI to complement and expand human capabilities is something we must get right.”
Adept was founded by a team with deep roots in large-scale neural networks. They have a tremendous amount of experience working with Transformers. This is not a sponsored post, but totally my own curiosity. This is the kind of startup that you’d want to be an early member of.
How could Adept foster AI-human workforce tools that could enable some glimmers of AGI? The following is also from Greylock’s blog:
A UNIVERSAL COLLABORATOR
Remember, while at Google, Ashish, Niki, and David trained bigger and bigger Transformers, with the goal of eventually building a model that could power all ML use cases. Along the way, the team discovered a major limitation: models like GPT-3 can write great prose, but they can’t take actions in the digital world.
Adept’s foundational and ambitious vision ties the general power of AI (and Transformer models) with advancements in program synthesis to enable AI models to both understand and act on behalf of users.
Adept’s general system will be trained on every software tool and API out there – leveraging the decades-long API-zation of all the tools and capabilities we use everyday to offer a universal collaborator that can work hand-in-hand with humans to radically improve effectiveness and productivity.
Adept will be able to do things like “submit an expense report for you” or “visualize a data set” – all using existing software like Salesforce, Figma, Tableau, and Twilio.
So they believe this product approach will play a central role in the future of work and transform many software markets. Most importantly, they also believe this is the most tractable definition of, and path to, AGI.
Let’s face it, this is the kind of startup that would be most likely to be acquired by someone like Microsoft, Snowflake, C3.AI, Salesforce or Google sometime in the future. It will be interesting to see if they get any traction trying to commercialize this, or to see exactly what it will be able to do.
Reid notes that he’s working with a range of key AI organizations, including OpenAI, Stanford Institute for Human-Centered Artificial Intelligence (HAI), Inflection AI, Microsoft, and now Adept AI Labs to try to get Human-AI augmentation right for the future of the workplace. The ties to Microsoft and its ecosystem are, in a sense, glaring.
Adept co-founders David Luan, Ashish Vaswani, and Niki Parmar -- certainly do have a broad experience including the Transformer. I would have wished the startup had more of a focus of an industry where they see their highest value enterprise customers coming from. The pitch of the product is super vague. Perhaps they themselves need time to figure out the product-market fit. When you think of how much Microsoft acquired Nuance for that was specifically for healthcare, you get an idea of how important this could become.
Users will work hand-in-hand with Adept’s technology, using a natural language interface to use existing software like Airtable, Photoshop, and ATS, Tableau, and Twilio. So this is also about how the AI-human hybrid workforce will use software better!
I think that’s what they are starting at Adept: the training a neural network to use every software tool and API in the world, building on the vast amount of existing capabilities that people have already created. This could become in some sense the “Stripe of the Neural Interface with Software” years from now.
So Adept is building a natural language interface to your computer – an AI collaborator for knowledge workers. That’s a bit like what we once hoped Cortana or Alexa at work would enable. These are certainly lofty ambitions!
A universal collaborator for every knowledge worker? What, pray tell, would that entail? Give me an AI-Assistant to help me work better.
Interesting in how the Great Automation companies will be very careful to remind us that they are augmenting people. While some AI companies are working towards a world where machines outperform humans, Adept believes machines should work together with humans. With Adept, people can spend more time doing work they enjoy the most.
Well, that’s a wrap guys. I’ll indeed be waiting for A.I. to get more Adept to be my work buddy and help me access software more seamlessly and automate tasks through simple voice queries.
I also expect the high-level collaborator to be a good student and highly coachable, becoming more helpful and aligned with every human interaction. But, developing AI to complement and expand human capabilities needs to get good first at a niche. A.I. needs to change the world one augmented employee or knowledge worker at a time!
What do you think about this?
The cost of an internship with cutting-edge AI systems is lower than it once was. With a pinch of OpenAI’s funding, recent startups, including AI21 Labs and Cohere, have created models comparative to GPT-3 in terms of their abilities. Many of us were thrown off by the pitch though. I think one quote that said it best is the following: “Adept’s technology communicates plausibly in theory, but talking about Transformers requiring to be ‘able to act’ feels like misdirection to me,” Mike Cook, an AI investigator at the Knives & Paintbrushes research collective, reportedly said.
Adept has a conception of the problem it wants to solve in the world, but needs some direction as to what its highest value customers really need it to be able to facilitate in the real-world. It’s time for the AI Lab researchers to get down to business.
Transformers might facilitate AI better suited to assisting in simple tasks. Just like RPI and no-code platforms, this will be ultimately good for society and productivity in the workplace.
Tl;dr Adept raised $65M from Saam Motamedi and Reid Hoffman at Greylock, Lee Fixel at Addition, Root Ventures, and angels including Scott Belsky (Founder of Behance), Howie Liu (Founder of Airtable), Dara Khosrowshahi (CEO of Uber), Chris Re (Stanford), Lip-Bu Tan (Chairman of Cadence), Andrej Karpathy (head of Tesla Autopilot), Nathan Benaich (Air Street), David Ha, Jaan Tallinn, Sarah Meyohas, and Kevin Kwok.
If you want to support me so I can keep writing, please don’t hesitate to give me tips, a paid subscription or some donation. With a conversion rate of less than two percent, this Newsletter exists mostly by the grace of my goodwill (passion for A.I.) & my own experience of material poverty as I try to pivot into the Creator Economy.
If you want to get my Creator updates, you can read the current issue here to hear what I am up to. You will notice on my home page of AiSupremacy, I have more interactive features on the right-hand margin.
Thanks for reading and have a good weekend! If you think I’m a creator worth investing in, give me a tip. It all goes to my basic needs and that of my family.
Join 66 other paid subscribers supporting me.