AI21 Labs, a Tel Aviv-based startup developing a suite of AI tools for text generation, has raised $155 million in a Series C funding round led by Walden Catalyst, Pitango, SCB10X, b2venture, Samsung Next, and Amnon Shashua, founder of Intel-owned Mobileye and co-founder For AI21 Labs. Google and Nvidia also participated.
This slide, which brings AI21 Labs’ total to $283 million, values the company at $1.4 billion — slightly higher than the reported figure. mentioned Earlier today ($1.2 billion).
“AI21 Labs came out of obscurity in October 2020 with our first launch of [AI writing tool] Wordtune,” Yoav Shoham, one of the co-CEOs at AI21 Labs, told TechCrunch via email. “Since then, we have been able to grow, doubling our workforce with plans to add more employees.”
AI21 Labs was founded in 2017 by Shashua, Shoham, and Uri Goshen, the startup’s other co-CEO. Shoham, a professor emeritus at Stanford University, previously sold two companies to Google, time management app Timeful and social network friend organizer Katango. Goshen is also a serial entrepreneur, having co-founded and led several technology companies in Israel, including communications analytics company Crowdx.
AI21 Labs’ flagship product is AI21 Studio, a pay-as-you-go platform for developers to create custom text-driven business applications from AI21’s proprietary text-generating AI models – including the cutting-edge Jurassic-2 model. The startup also sells access to the aforementioned Wordtune, a multilingual reading and writing AI assistant similar to Grammarly.
Customers can leverage the AI21 Labs platform via APIs for specific AI use cases, such as summarizing, paraphrasing, and grammar and spelling correction. The startup’s models support a growing number of languages, including Spanish, German, Italian and Dutch.
AI21 Labs competes with an expanding list of startups and established companies in the exciting field of generative artificial intelligence.
Google, AWS, and Microsoft offer tools similar to AI21 Studio, as do startups like Cohere, OpenAI, and Anthropic (and to a lesser extent marketing-focused vendors like Jasper, Regie, and Typeface). AI21 laboratories suffer from a lack of funding; OpenAI has raised $11.3 billion to date, while Anthropic and Cohere have raised $1.6 billion and $435 million, respectively.
But Shoham stresses that AI21 Labs’ solutions are superior in several aspects, even though they look similar on the surface and don’t have the advantage of a higher R&D budget.
For example, Shoham says, it was developed on “some of the largest and most sophisticated big language models in the world” and provides “more fine-grained control” than many generative AI applications on the market. Moreover, they are trained on recent data, unlike text generation models trained on old data, which cannot accurately answer questions about current events.
AI21 Labs offers a range of tuning parameters to customize the output of its language models, ranging from small, fast models to larger, more sophisticated displays. Image credits: AI21 Labs
“AI21’s AI systems are easy to integrate and generate reliable, trustworthy and accurate results,” Shoham said. “The quality and quantity of training data is a huge challenge in the industry. We have done our best to ensure that we are using up-to-date data, constantly training our models on additional data and working to ensure that the data used is reliable and trustworthy.
Since I haven’t tested AI21 Labs products recently, I can’t speak to the veracity of these claims. But whether it’s hyped or not, AI21 Labs seems to be gaining some attention despite its flaws.
Wordtune alone has more than 10 million users, says Shoham — who wouldn’t disclose the size of AI21 Labs’ customer base, other than that it includes “several” Fortune 100 companies — AI21 Labs was a launch partner of Bedrock, Amazon’s generative AI application development platform.
Shoham says the latest funding tranche will allow AI21 Labs to accelerate its R&D efforts and “reach its goal of developing the next level of AI,” with “thinking capabilities across many domains.” AI21 Labs also plans to pursue more partnerships with companies “across the entire technology ecosystem” and increase its headcount of 200 people, focusing primarily on roles in research and business development.
AI21 Labs continues to build out its executive suite as well, recently hiring Pankaj Dugar, former Google and Databricks executive, as senior vice president of marketing and general manager of North America.
You will definitely need money, given the capital-intensive nature of developing large language models.
according According to The Information, OpenAI spent $540 million last year to create ChatGPT. and its AI21 Labs research It pegs the expense of training a text-generating model using 1.5 billion parameters (i.e., the variables the model uses to generate and analyze text) at $1.6 million. For reference, AI21 Labs’ Jurassic-2 model’s predecessor, Jurassic-1, has 178 billion parameters.
This does not take into account hosting costs for the templates service.
OpenAI was at one point It said Paying up to $700,000 per day to keep the infrastructure hosting ChatGPT running. Maths on the back of the napkin Pegs The cost of running a model the size of GPT-3, and the older model developed by OpenAI, is $87,000 per year on a service like AWS.
AWS is a partner of AI21 Labs, so it’s not clear if there’s some sort of downsizing in play. But I’d bet to say that this overhead is not insignificant.
“We have focused on growing responsibly, which allows us to continue to grow and expand, providing our customers with the latest and most trustworthy AI technologies,” Shoham said.