American tech startup Cerebras Systems has partnered with Cirrascale Cloud Services to make generative AI technology more inclusive.
The partnership focuses on developing a Large Language Model (LLM), a generative AI tool that can mimic human communication. It can read, outline, translate and even predict future texts. According to Cerebras CEO Andrew Feldman, LLM does not receive the hype it is supposed to have. He insisted, however, that it would soon change.
Within the next year, we will see a sweeping rise in the impact
Andrew Feldman, CEO
“Within the next year, we will see a sweeping rise in the impact of LLMs in various parts of the economy,” Feldman said.
The brand new Cerebras AI Model Studio enables users to train generative Transformer (GPT)-class models — such as GPT-NeoX and GPT-J — at lower costs than competitors by utilizing a simple code. The Cirrascale AI Innovation Cloud hosts this program, while the LLM training occurs on Cerebras Wafer-Scale Clusters.
In the Cerebras AI Model Studio, users can choose their preferred GPT-class model, which ranges from 1.3 billion to 175 billion parameters. LLM training can be done within a shorter period, resulting in low costs. Feldman explained that it was made possible by the innovative cloud technology used in this tool.
Most traditional cloud providers cannot guarantee variable latency between large numbers of GPUs, increasing the resources necessary to complete training. With Cerebras, training duration can take only eight days as opposed to the usual 64 days. On traditional clouds, the GPU production cost alone can amount to $61,000. Meanwhile, Cerebras’ whole production cost is only around $45,000.
The Cerebra AI Model Studio allows users to forgo DevOps and distributed programming because its push-button scanning can run up to 20 billion parameters. Users can also train their models with longer sequence lengths, which allows more development opportunities.
In addition to the partnership with Cirrascale, Cerebras also announced a collaboration with content platform Jasper AI to accelerate the adoption of generative AI technology and enhance its reliability.
AI News Content
In the past few years, AI content tools such as Rytr and Jasper have proliferated with the release of Open AI GPT-3. Another player in this domain is Monok, with their own AI solution for curating and generating news articles to increase organic traffic.
One thing that sets Monok apart from their competitors is that their subscription model ensures customers always receive value for their monthly fee, unlike Rytr and Jasper where remaining monthly credit is not passed on to the next month. Monok also utilizes human editors for curation of generated stories and editing of generated articles to ensure highest level of text quality.
