Thousands of tech industry people gathered at INNOVATEwest in Vancouver this week
Tech entrepreneurs visiting Vancouver this week received a crash course in the rapidly evolving world of artificial intelligence (AI).
Experts from technology companies and institutions participated in a panel discussion at the INNOVATEwest conference to share best practices for startups and scale-ups when implementing AI. The conference, held April 16th and 17th at the downtown convention center, attracted thousands of participants.
Panelists urged startups to reconsider investing in training their own models and start with existing tools to control costs while exploring the benefits of technology.
“Don't start by training your own model. That's the wrong way to go. You'll have to pay freemium or premium fees, and you'll have to build many prototypes with millions of tokens and contexts. , to be able to prove something and build our first product,” said co-owner Ryan Monsurate.The Vancouver-based founder and chief technology officer of Farpoint Technologies Inc., an AI consulting firm.
Ryan Nadel, principal product manager at Microsoft (Nasdaq:MSFT), says it's important for companies not to start with an optimization mindset when building large-scale language models (LLMs) and other types of models. said.
“If you look at the trajectory of technology history, the way you build great things is to build things and then adjust, tweak, and optimize it to reduce costs and find margins,” Monserate said. said.
“I highly recommend not optimizing too quickly, focusing on quality and value, and then focusing on profit and COGS. [cost of goods]”
As many companies use ChatGPT and other AI tools to assist with daily operations, Monsurate needs to provide the tools with context about the company and the project, such as how to onboard and orient new employees. I said that there is. result.
“I tend to use audio input a lot. The input in ChatGPT Plus is very good, and I can just talk to it for five minutes and say, 'Here's all the other information I can give you about this issue.' can do. Now try to solve this problem. That way you will get better results. ” he said.
Monsurate added that companies should check their user license agreements to see if machines are capturing company information to improve models. Typically, the enterprise version does not allow you to legally use company information, and the personal version gives you a certain amount of time to delete your chat history.
And Monsurate said ChatGPT isn't the best use case for every task. Companies often want to merge his ChatGPT with other types of models, such as vision-based models or audio-based models. In some cases, you may be able to accomplish some of your more complex goals by merging all of these models into one larger model. .
Bethany Edmunds, assistant dean of Northeastern University's Computing Program, said that when leveraging AI, companies can adapt to their needs to not only provide accurate information but also be “the best version of themselves.” There is a need to ensure that there is a process in place to revise the model. Vancouver campus.
“A lot of the elements are trained on the internet, but that's not necessarily what we want to recreate,” she said.
“So when mistakes are made, who is to blame? Who looks for bias? Who considers their impact? It’s really important to embed that throughout the organization. If that’s a software engineer’s job, you need time to do it.”
When it comes to AI talent, there aren't enough experienced workers to meet the demand at the moment, so Monsurate said it's hiring interns fresh out of college and training them on the job.
“I think it's very valuable for all startups to consider… It's a simple thing that people might be willing to do, and it's a two-bird [with] This is because finding a specialist takes a lot of time and money. ”
[email protected]
