Six more companies competing with OpenAI
Meet six companies carving out their own space in the AI industry with their unique approaches and applications of generative AI 🔮
One of my most popular guest posts to date covered OpenAI’s main competitors, this time around I asked very talented software developer Conrad Gray to help me cover many others. I spotted Conrad’s work on Substack and immediately realized he’d be the perfect person to tell their story. Check out his Newsletter:
Conrad Gray is a software engineer who, after almost two decades of writing code, decided to become a bioengineer. On H+ Weekly, he sheds light on the bleeding edges of technology and discuss how advancements in AI, robotics, and biotech can bring about abundance, expand humanity's horizons, and redefine what it means to be human. Some analysts and sci-fi enthusiasts refer to this as transhumanism. His Newsletter is actually eight years old.
To read more by Conrad Gray, check out these pieces:
For the latest info on A.I. trends, become a premium subscriber of A.I. Supremacy.
Let’s dive right in:
Six more companies competing with OpenAI
Meet six companies carving out their own space in the AI industry with their unique approaches and applications of generative AI.
In the previous post in this series, we explored some of the major competitors of OpenAI. With the rapid growth of the AI industry, numerous startups have emerged in recent years. In this article, we will meet six more AI startups that are carving their own space in this fast-paced industry.
If Gmail says this Email is too long, go here.
Founded in 2022 by ex-Google and ex-DeepMind engineers
An action-oriented AI assistant designed to perform complex tasks from a text prompt
Adept AI was founded in January 2022 by three AI researchers - David Luan, Ashish Vaswani and Niki Parmar. Two of the founders, Vaswani and Parmar, may be familiar to some of you as they were among the authors of the influential Attention is All You Need paper. This paper introduced the concept of transformers in AI research. Without them, large language models such as GPT-4 or PaLM would not exist and the current AI revolution would not happen.
Adept AI emerged from stealth in April 2022 and shared what they are working on - an AI assistant, or “a universal collaborator for every knowledge worker” as they have described it, capable of doing almost anything a human in front of a computer can do. Alongside this announcement, Adept AI revealed a $65M Series A funding round led by Greylock, Addition, and Root Ventures, as well as introduced the founding team consisting of ex-Google and ex-Deep Mind engineers and researchers.
Nothing concrete was known about the product until about six months later when the team showed their own large language model - ACT-1. As Adept AI writes, ACT-1 is an Action Transformer, specifically designed to take actions.
ACT-1 looks like a popup in your browser where you type what you want it to do and it will do it. With ACT-1, beginners can become power users. “Anyone who can articulate their ideas in language can implement them, regardless of expertise”, the company promises.
Examples shared by Adept AI demonstrate ACT-1's capability to navigate websites and perform various tasks, ranging from simple spreadsheet operations to more complex activities like finding a home for sale based on specific criteria.
It seems ACT-1 is able to break a complex task into a series of smaller steps, similar to what open-source AI researchers and enthusiasts have done with AutoGPTs.
At the beginning of this year, Adept's two co-founders, Ashish Vaswani and Niki Parmar, left the company. Vaswani held the position of Chief Scientist, while Parmar served as CTO at the time of their departure. The official reason for their exit has not been disclosed. According to their LinkedIn profiles, both Vaswani and Parmar are currently involved in a stealth startup.
In March of this year, Adept successfully concluded a $350M Series B funding round, led by General Catalyst and co-led by Spark Capital. The exact valuation of the company remains undisclosed to the public, but there are hints from Reuters suggesting that Adept AI may already be valued at over $1B.
Founded in 2022 by Karén Simonyan (former AI researcher at DeepMind), Reid Hoffman (co-founder of LinkedIn), Mustafa Suleyman (co-founder and former head of applied AI at DeepMind)
An AI chatbot focused on user experience
Many chatbots claim to be personal assistants, but no company has embraced this concept as fully as Infection AI. Founded in 2022 by Karén Simonyan (former AI researcher at DeepMind), Reid Hoffman (co-founder of LinkedIn), Mustafa Suleyman (co-founder and former head of applied AI at DeepMind), Inflection AI is on a mission to create a true personal assistant that can be your coach, confidante, creative partner or sounding board.
“It feels like we’re on the cusp of being able to generate language to pretty much human-level performance. It opens up a whole new suite of things that we can do in the product space”, said Mustafa Suleyman, as quoted by TechCrunch.
To fulfil this mission, Inflection secured a $225M Series A funding round and assembled a team consisting of experienced AI researchers who previously worked at such places as DeepMind, Google, Meta, Microsoft and OpenAI.
The company came out from stealth mode in May 2023 with the announcement of Pi. Pi (which stands for “Personal Intelligence”) is designed to be a kind and helpful companion. This is reinforced by the visual design of the chatbot. It’s clean and radiates calmness. Pi is the most aesthetically pleasing chatbot out there.
Example conversation with Pi
Pi is powered by a proprietary large language model developed and trained by Inflection AI. It is not using any third-party models like GPT-4. The team behind it put a considerable amount of attention to giving Pi a friendly and fun personality.
Another thing that makes Inflection AI unique in the new wave of AI startups is that they are a Public Benefit Company (also known as B Corporation). As Inflection states, their purpose is to “develop products and technologies that harness the power of AI to improve human well-being and productivity, whilst respecting individual freedoms, working for the common good and ensuring our products widely benefit current and future generations”.
As of the time of writing, little information is available regarding Inflection's monetization strategy for its AI. Currently, Pi is free to use, and there are vague plans to introduce paid subscriptions. One possibility is that Inflection intends to generate revenue through its Conversational API, which is currently available via a waitlist.
Editor’s Note: Try HeyPi Chatbot here. Podcast: Personal Intelligence with Mustafa Suleyman [AI miniseries]
Give A.I. Supremacy a Testimonial
If you enjoy this Newsletter, help us build some social proof.
Founded in 2018
AI Magic Tools - a suite of video editing AI tools
Text-to-image, image-to-image, text-to-video and video-to-video AI generators
One of the promises of generative AI is the democratization of creativity. With tools such as Midjourney or DALL-E, you can create artwork exactly as you imagine them without spending time learning how to make them. One of the companies spearheading this application of AI is Runway.
In fact, Runway played a significant role in kickstarting this movement. Patrick Esser, Principal Research Scientist at Runway, is a co-author of High-Resolution Image Synthesis with Latent Diffusion Models. This paper introduced the concept of Stable Diffusion, which powers many text-to-image generators, including Midjourney.
Creativity is deeply ingrained in Runway's DNA. The founding trio—Anastasis Germanidis, Alejandro Matamala-Ortiz, and Cristóbal Valenzuela—met at art school at New York University. In 2018, they established Runway as an applied AI research company with a mission to build the next generation of creativity tools and make them accessible to all creatives. Armed with $2M seed funding from Lux Capital, Runway set out to transform the creative work.
Runway offers a suite of AI-powered tools for creatives. These AI Magic Tools can expand an image, interpolate frames, remove objects from videos, turn one image into another or generate a new one from a text prompt.
See examples here.
AI Magic Tools also includes two powerful video generators - Gen-1 and Gen-2.
Gen-1 takes a video as input and transforms it into the desired style. With a single prompt, a scene shot on an iPhone can be turned into claymation, anime, or any other style.
See examples here.
Gen-1 video-to-video generator in action. Source: Runway
Gen-2 is a text-to-video generator. It is similar to ChatGPT or Midjourney, but instead of text or images, it generates videos based on text descriptions. Users can describe the desired scene, and the AI generates the corresponding video.
Example videos generated by Gen-2. Source: Runway
Runway's tools have been utilized by organizations such as CBS, Nickelodeon, and Vox, as well as popular YouTubers like Mr Beast and Corridor Digital. Their tools have been used on CBS's Late Show with Stephen Colbert and in the production of "Everything Everywhere All at Once" and "Top Gear America."
Runway's AI Magic Tools have the potential to revolutionize the filmmaking process and provide powerful editing tools to a wider audience. In the early 2010s, accessible and powerful editing tools like Adobe Premiere and After Effects, coupled with the declining prices of digital cameras and the rise of YouTube, led to an explosion of creativity on the internet. A similar explosion may occur with the generative AI tools offered by Runway. Traditionally, a film crew would need to rent an expensive green screen studio and a motion capture system, and then spend hours in post-processing to achieve the director's vision. With AI Magic Tools, anyone with a smartphone can replace any background with another and easily alter actors and scenes at a fraction of the cost and in less time.
Runway recognizes the potential of these tools and puts effort into nurturing the next generation of storytellers enhanced by powerful AI tools. Runway Studios has been launched with this goal in mind, offering grants to aspiring filmmakers from diverse backgrounds in need of production support.
Earlier this year, Runway hosted the first annual AI movie festival, the AI Film Festival in New York, to showcase the possibilities of creativity enhanced by AI tools. The festival winners' works can be explored on the website.
Founded in 2019
Luminous - an AI platform with explainability tools designed for European market powered by multimodal multilanguage large language model
Aleph Alpha was founded in 2019 in Heidelberg, Germany, with the mission to become the leading European AI research company and ensure European digital sovereignty. To fulfil this mission, Aleph Alpha has raised €28.3M ($30.4M) up to date.
Jonas Andrulis, CEO and founder of Aleph Alpha told TechCrunch: “Aleph Alpha’s mission is to enable the accessibility, usability and integration of large, European multilanguage and multimodal AI models following the likes of GPT-3 and DALL-E, driving innovation for the explainability, alignment and integration. The funds significantly accelerate the process of bringing the latest generation of AI technology into application and secure the digital sovereignty for public and private sector partners in Europe and beyond.”
Aleph Alpha’s primary product is Luminous, an API platform powered by a large language model that enables other companies to build their own AI solutions. It is available in four variants: Base, Extended, Supreme, and Control Models. All models can take text as input. Only Luminous-base and Luminous-extended are multimodal, which means they can accept both text and images as inputs.
Luminous models. Source: Aleph Alpha API documentation
All Luminous models have been trained in the five most commonly spoken European languages: English, German, French, Italian, and Spanish.
Luminous comes with tools to better understand how the model works. These tools allow developers and researchers to suppress or amplify specific aspects of the model and investigate the reasons behind its behaviour. They can also help detect and investigate hallucinations.
The explainability tools are a unique feature designed by Aleph Alpha specifically for European customers. With the upcoming EU AI Act, transparency and traceability will be required for AI models, and Luminous is already prepared for these regulations.
As part of the platform, Aleph Alpha also offers access to a conversational model named Lumi. Internal benchmarks published by Aleph Alpha suggest Luminous is performing similarly to OpenAI’s GPT-3.
Editor’s Note: A German startup.
Founded in 2017
Wordtune - AI-writing assistant that catches grammatical errors and rephrases the text
AI21 Studio - an AI platform focused on providing access to powerful language models for developers
Based in Tel Aviv, Israel, AI21 Labs was founded in 2017 by Yoav Shoham, Ori Goshen, and Amnon Shashua. The company is focused on written language and how reading and writing can be enhanced with artificial intelligence. Their mission is to reimagine “the way humans read and write, for the better”.
AI21 Labs’ first commercial product was Wordtune, released in 2020. Wordtune is an AI-writing assistant, similar to Grammarly, that catches grammatical errors and rephrases the text to match different styles, like making it sound more casual or more professional. Wordtune was a successful product and become one of the favourite Google Chrome extensions for 2021.
In August 2021, AI21 Labs announced AI21 Studio - an AI platform focused on providing access to powerful language models for developers. It was powered by Jurassic-1, their homebrew large language model with 178B parameters (for comparison, OpenAI’s GPT-3 has 175B parameters). With AI21 Studio, developers can easily build applications utilising powerful text operations.
In March 2023, AI21 Labs announced Jurassic-2, their latest family of large language models. According to AI21 Labs, Jurassic-2 improves upon Jurassic-1 in every way and adds support for several non-English languages, including Spanish, French, German, Portuguese, Italian and Dutch.
Alongside the new model, AI21 Labs introduced a suite of new Wordtune APIs for text operations such as summarization or paraphrasing. According to AI21 Labs, their offering is better and faster than OpenAI APIs.
AI21 Labs’ models and APIs are also available on Amazon Bedrock - AWS’s suite of APIs and models for building applications using generative AI.
Editor’s Note: An Israeli Startup.
Beijing Academy of A.I.
Founded in 2018
State funded, non-profit research company
Wu Dao 2.0 - the largest known multimodal large language model
In 2017, China's State Council released a document outlining an ambitious plan to have China cause “fundamental breakthroughs by 2025” in AI research and to become “the world’s primary innovation centre by 2030” (the English translation can be found here).
To fulfill this vision, the Beijing Academy of Artificial Intelligence (BAAI) was established. Founded in November 2018, this government-sponsored non-profit research lab is tasked with advancing AI research in China. It aims to promote collaboration between academia and industry, nurture top talent, and focus on long-term research on the fundamentals of AI technology.
In 2020, BAAI began developing its own large language model, leading to the release of Wu Dao in January 2021. Just six months later, in May, BAAI unveiled Wu Dao 2.0, a massive model with 1.75 trillion parameters, making it the largest model ever released. For comparison, GPT-3 has 175 billion parameters, making Wu Dao 2.0 ten times larger.
Wu Dao 2.0 is a multi-modal AI system capable of responding to prompts, performing image recognition, and generating images based on prompts. According to Engadget, Wu Dao can also be used to power virtual idols and predict the 3D structures of proteins, similar to AlphaFold.
To showcase the capabilities of Wu Dao 2.0, BAAI, in collaboration with Zhipu.AI and Xiaoice, created Hua Zhibing, China's first virtual student attending Tsinghua University, one of the country's top universities.
In a video shared on Weibo, Hua Zhibing introduced herself and presented a poem and painting she generated. The background music was also composed by the AI. (note: the video itself has not been generated by an AI. From what I found out in my research, the girl in the video is a real person and only her face has been swapped with an AI-generated face).
“The way to artificial general intelligence is big models and big computer,” Dr. Zhang Hongjiang, chairman of BAAI, said to Engadget during the conference at which Wu Dao 2.0 was unveiled to the world. “What we are building is a power plant for the future of AI, with mega data, mega computing power, and mega models, we can transform data to fuel the AI applications of the future.”
In addition to Wu Dao, BAAI is also working on the Jiuding AI computing platform, MetaWorm 1.0 (Tian Bao), which recreates the nervous system of the C. elegans nematode, and other projects aimed at advancing AI research.
Read more by this author
Unlock free complimentary subscriptions while sharing the Newsletter with a co-worker, friend or acquaintance.
Poll on Promising A.I. Startups
Let me know who else you want me to cover?
Last time we covered:
I also host interactive Chats on A.I.’s impact on society here.
Thanks for reading!