How companies are experimenting with services like ChatGPT

Tool makers in the knowledge economy are embracing the innovation frenzy more fully. Microsoft has announced a series of product updates that help desk jockeys ease tasks from drafting emails and summarizing documents to writing computer code. “It’s like working in the year of the dog,” said Eric Boyd, his head of AI for the company’s cloud computing division, describing the company’s hectic release schedule. His rival Google, like Adobe, is beefing up its tool suite. Salesforce and Bloomberg are building software for creative types, salesmen and financial gurus respectively, and startups like ChatGPT-like legal assistant Harvey and writing aid Jasper are on the rise. .

Despite all the experimentation, companies are still unsure how to take advantage of AI’s emerging capabilities. Boyd said most people either underestimate or overestimate the capabilities of this technology. Efforts are being made to determine which professions are the most likely candidates for reinvention. A study released last month by his OpenAI, the organization behind ChatGPT and DALL-E, looked at the proportion of tasks within occupations that could be speeded up by at least half using the new technology. High on the list were occupations that involved a large amount of routine writing, number crunching, and computer programming, such as paralegals, financial analysts, and web designers.

It is unlikely that companies will eliminate such jobs entirely any time soon. Generative AI works well for creating first drafts, but relies on humans to provide direction and evaluate results. You could say that Microsoft labels its new suite of tools as “co-piloted.” Social for Professionals Co-founder of his LinkedIn, Reed, in his Hoffman’s recent book “Impromptu,” the author advises users to deal with her ChatGPT and more. “It’s like an undergraduate research assistant.” (This book was written with the help of bots.)

What’s more, there’s still little evidence that companies are reducing the numbers of programmers, salespeople and other white-collar workers as they become more productive, argues Michael Chewie of consultancy McKinsey. Software may eventually eat up the world, as one venture capitalist predicted, but so far it’s only scratching the surface. And most businesses would certainly choose more sales over fewer salespeople. However, there are many hurdles ahead for companies looking to leverage generative AI. First, many companies will need to rethink the role of junior staff as apprentices to be trained, not workhorses to be whipped. Getting the most out of generative AI can be challenging for companies with clunky old IT systems and scattered datasets. On the plus side, large-scale language models like the one powering ChatGPT are better at manipulating unstructured datasets than previous types of his AI, he said, according to a consultancy firm partnered with OpenAI. Bain’s Roy Singh said:

Other reservations may still delay adoption. Will Grannis, chief technologist for Google’s cloud computing division, said the hurdles for adopting new technology are much higher for businesses than for consumers. One concern is the protection of sensitive data, with companies from bank JPMorgan Chase to defense contractor Northrop Grumman banning employees from using ChatGPT at work. Zurich does not allow personal customer information to be entered into the tool.

An even bigger concern is reliability. Tools like ChatGPT can spew plausible but inaccurate information. This process is euphemistically called “hallucination”. While this may not be a problem when creating promotional materials, it is a fatal flaw in other areas. It’s an airplane wing,” said Mike Haley, head of engineering at his software maker Autodesk. Humans make mistakes too. The difference is that, for now, generative AI tools don’t explain their mindset or confess their level of confidence. This makes it difficult to trust when the risk is high.

Productivity for people

Bosses may also undermine their appetite for generative AI, especially as technology becomes smarter, with growing concern about the risks it poses to society. Some worry about the barrage of scams, misinformation, and computer viruses that AI creates. These concerns are prompting government action. The US Department of Commerce is seeking public comment on how the technology should be approached. The European Union is amending its planned AI bill to encompass recent advances. Italy currently bans ChatGPT.

A final concern is that smart AI can demoralize staff who worry about the future. But so far, employees seem to be among the most enthusiastic supporters of the new technology. Of the 12,000 employees surveyed by his workplace network app, Fishbowl, in January, 43% used tools such as his ChatGPT for work-related tasks, the vast majority without their bosses knowing. was used for Such enthusiasm suggests that few people shed tears when simple tasks are taken over by AI. “Nobody goes to law school to spend time digging through paperwork,” said Winston Weinberg, co-founder of Harvey. That might be enough to motivate companies to keep experimenting, but it won’t be possible after two decades of slow productivity growth in rich countries. bad thing.