[ad_1]
Large machine learning models are at the heart of technologies such as OpenAI’s DALL-E 2 and Google’s LaMDA. Those who can generate enough images and text to pass on to people are certainly amazing. But it took a lot of time to prepare the models and calculate the power – not to mention the money. DALL-E 2 alone trained for 256 GPUs for two weeks, costing about $ 130,000, according to one source. Estimate.
Small companies find it difficult to keep up, which is why many turn to AI-as-a-service providers who control the challenge of creating models and charge them through the API. One such provider is AssemblyAI, which specializes in speech-to-text and text analysis services.
AssemblyAI announced today that it has raised $ 30 million with participation in a series B led by Insight Partners Y Combinator and Stripe co-founders John and Patrick Collison, Nat Freddman and Daniel Gross. To this day, AssemblyAI has raised $ 64 million, which founder and CEO Dylan Fox is investing in to improve the company’s AI model training for Tech Cruises and Data Center.
After a two-year stay in Cisco, Fox established AssemblyAI, working on machine learning of collaborative products. Prior to that, Yugiv 1 started working with companies to reward its customers with non-profit donations.
“I was looking for Speech Recognition and Natural Language Management (NLP) APIs for previous projects, and I started AssemblyAI after seeing how limited and low-accuracy options were back in 2017,” Fox told TechCrunch in an email interview. “The company’s mission is to explore and deploy the best AI models for NLP and speech recognition and to expose those models to developers with free and easy-to-integrate software development tools and APIs.”
AssemblyAI provides AI-powered, API-based services for automatic summary, title search and content editing in more than 80 languages, and divides audio and video files into “chapters” for each summary. Using the platform, developers call various APIs to perform tasks such as “Identify speakers in this conversation” or “Check this podcast for restricted content” starting at $ 0.00025 per second at a relatively low cost.
Image thanks AssemblyAI
“We’re training giant AI models on billions of GPUs on billions of units,” Fox said. “Parameters” indicate the size of the models; In general, larger models are more sophisticated. ”Using advances in AI research, we will continue to dramatically improve the accuracy of all our AI models and launch new ones. “Ohur ‘AutoTrain’ feature allows the API to learn random data from client data to automatically update over time.
AssemblyAI is not the only player in the congested AI-as-service sector. NLPCloud provides NLP models out of the box with APIs, and Sayso created an API to convert English language from one accent to real-time. Not in vain, Amazon, Google and Microsoft have a host of API-based products that target applications such as text analysis, image recognition, text-to-speech, speech-to-text, and more.
But Fox says AssemblyAI continues to grow rapidly The epidemic, and – by extension – the rise of distance work. Audio and video are merging with expanding products, such as video conferencing and dating apps. That is what leads PHold groups to find ways to build more, more valuable features on audio and video data.
“These features are the tools of content editing or advertising platforms for collaborative tools that identify topics in podcasts and videos, read transcripts, summaries, and keywords for video messages shared on platforms. “They are building smart communication centers and revenue information products that can analyze support and sales calls,” Fox said.AssemblyAI is fast becoming the go-to API platform so that these product teams can send these AI-enabled features to the audio and video data in their products.
According to Fox, AssemblyAI now has “hundreds” of customers paying more than 10,000 users. Since the beginning of 2022, the user base has increased 3x, while revenue – which Fox refused to disclose – has reached 3x.
”[We’re] Making millions of API calls every day, ”said Fox. “We plan to triple our AI research team over the next six months and invest millions of dollars in GPU hardware to train larger and more complex AI models. This pushes the envelope.
Fox believes the growth will make AssemblyAI better next year – they can make any headway. At a time when redundancies are becoming more common and difficult to finance, AssemblyAI said it will double its 52-member group by the end of the year.
“A few months ago, we went into our S Series A, which we closed from Asil in February, and we didn’t have to actively fundraise. But we met Rebekah. [Liu-Doyle] For a while from Insight, and she, Insight, with her big and extra capital, really felt she could help us. [spur] We are making more progress, ”said Fox. “AWhen the market opens, we need to be able to establish ourselves as a leading supplier in this space and support our growing customer base – more accurate AI models can support the features and products you are building.
[ad_2]
Source link
