Want Smarter Spights in your Inbox? Sign up for our weekly newsletters to get what items on business leaders, data, and security leaders. Subscribe now
Deciding on AI models is most of a technical decision and it is a strategic way. But choose open, closed or hybrid models all have trade-offs.
While speaking this year change VB, model architectural experts from General Motors, Zoom and IBM mentioned how their companies and customers consider the AI model.
Barak Turovsky, which is in March becomes the first GM officer, said there are many noises on each new model resease and every time the leaderboard changes. Long before leaderboards are a mainstream debate, Thotovsky helped the first Many language models .
“It’s probably one of the biggest breakthroughs that helped Openi and others begin to launch,” Thamovsky said. “So honestly a funny anecddote: the open source helped make something closed and now need to be open.”
The reasons for decisions vary and include cost, performance, trust and safety. Turovsky said businesses sometimes want a mixture of strategy – using an open model for internal use and a closed model for facing production and vice versa.
IBM’s Strategy
Armand Ruiz, the VP of the IBM’s AI platform, says IBM started its platform own llmsBut then it knows it is not enough – especially because the most powerful models arrive in the market. The company has then been extended to offer equilibrium with platforms such as clutching face so customers can choose any open source model. (The company has recently drawn a new model GATES that gives businesses an API for moving between LLMS.)
Many businesses are chosen to buy more models from many vendors. If Andreessen Horowitz examined 100 cios, 37% of respondents said they used 5 or more models. Last year, only 29% use the same amount.
Choosing is the key, but sometimes the extra option creates confusion, says Ruiz. To help customers in their way, IBM does not worry about which LLM they used during the testimony of the concept or pilot phase; The main purpose is the possibility. Only later they began to see if there was a model or based one of the needs of a customer.
“First we try to simplify all paralysis of analyzing all options and focus on the case of use,” Ruiz said. “Then we know what’s the best way for making.”
How Zoom Ai is coming
Zoom’s customers can choose between two configurations for AI companion, as Zoom CTO Xued Huang. One involves Federate own LLM with other larger foundation models. Another delivery allows customers concerning the use of many models that only zoom model is available. .
The company makes his own little language model (SLM) without using customer data, says Huang. In 2 billion parameters, LLM is actually very small, but it can still be above other specified industries. SLMs are best in complex tasks when working with a larger model.
“This is the power of a hybrid approach,” said Huang. “Our philosophy is very straightforward. Our company leads the road like Mickey Mouse and the small model to be a small model to be a small model to be a small model to be a small model to be a small model. The smaller model can be a small model. model to be a small model to be a small model. The small model can be a small model to be a small model to be a small model to be a small model to be a small model to be a small model to be a small model to be a small model A small model to be a small model to be a small model to be a small model. The small model can be a small model to be a small model. A small model together.
