This article originally appeared on Bloomberg News on 1 May 2023.

One area where AI can have a significant impact is in optimizing the growing process. By analyzing data on plant growth and environmental conditions, AI algorithms can help growers identify optimal conditions for maximizing yields and potency.

Psych! That was one paragraph of ChatGPT’s four-graph answer to the question. And it wasn’t half-bad. But cannabis’s interaction with artificial intelligence will be much more complex than that. Here’s another stab at the question, from a human author:

PotGPT

Zeta Ceti, the founder of a consulting firm that has helped cannabis entrepreneurs get more than 100 licenses in two countries, had used human consultants for more than a decade. And then ChatGPT inspired his company, Green Rush Consulting, to launch a chatbot specific to the marijuana industry.

Oddysee AI, which Ceti demonstrated at a marijuana conference in midtown Manhattan last week, is designed to do around 70% of the work required to navigate cannabis licensing. The other 30% will still be done by human employees, Ceti said, adding that the technology preserves the consulting firm’s intellectual property.

“Six people built it in 33 days,” said Ceti, while standing outside the company’s green and black booth at MJ Unpacked. The self-funded project is currently looking for investors, according to Ceti, who named himself after a binary star.

Oddysee AI isn’t the only startup in cannabis AI. Little Dragon, a delivery-only business that specializes in high-THC products called dabs, also launched an AI program recently. And Predictmedix, a Toronto-based company, is developing a program to test for cannabis impairment.

As the cannabis industry jumps on the AI bandwagon, the results are likely to vary widely, considering there are different regulations and issues in each of the more than 35 states that allow some form of licensed cannabis.

Oddysee AI co-founder Jason Rosenberg, asked to explain his product’s usefulness, said his chatbot will respond to a question about Cookies by directing users to the marijuana brand by that name, founded by the rapper Berner. “If you go to ChatGPT and ask it, you’re more likely to get something about Oreos,” he said.

“You have ChatGPT and Bard. But no one’s niching it out. You don’t want to have to ask 20 questions to get to where you need to go,” Rosenberg told me.

Little Dragon was launched on TruCrowd’s equity crowdfunding platform. Co-founder Rachel Figueras said in a promotional video that the AI can learn a user’s preferences and help them select products better than a human budtender can.

Such initiatives could put more than a few budtenders and licensing consultants out of business. They also raise questions about what could happen to the industry if some issues are tackled before others. After all, the cannabis business has grown haphazardly, state by state, and has worked on the question of how to sell its products without first solving overarching issues such as standardization, health risks or impaired driving.

And that’s become its biggest problem. States are slowly inching along on parallel paths to try to solve many of the same problems. Washington, Vermont and Colorado, for example, have all considered bills seeking to cap high-potency THC such as the products that Little Dragon would flavor-match and deliver to consumers’ doors. California’s overabundance of cannabis companies are struggling to consolidate — a painful process that could be exacerbated by AI-assisted entrepreneurs getting rapid access to licenses.

Impaired driving is another hurdle that’s hamstrung hesitant states. (Note to reader: Just in case you doubted a human was writing this, ChatGPT doesn’t alliterate.) Many state politicians have cited the lack of an effective test for whether someone is too impaired by cannabis to operate a vehicle  as a reason to hold off legalizing. AI solutions are stepping in to try to solve that challenge.

Predictmedix, which trades on the Canadian Stock Exchange and the Frankfurt Stock Exchange, said it uses multispectral cameras to analyze physiological data patterns and identify any kind of impairment, regardless of whether it comes from alcohol, cannabis, mental illness — or even a condition like Parkinson’s.

“AI-based impairment testing can be easily scaled up to accommodate large numbers of people, making it useful for mass-testing scenarios such as roadside checkpoints or workplace screenings,” Chief Operating Officer Rahul Kushwah said in an emailed statement. Pilots have already been carried out for events like Formula 1 races, he noted.

So, as it embraces AI, what problem should the cannabis industry try to solve first?

I asked the expert.

“The need for standardized testing and labeling,” answered ChatGPT. Coming second, according to the program, was the need for “more research and scientific data on the effects of cannabis.”

Unfortunately for the industry, the US federal government is tackling those issues in the opposite order. President Biden made it clear that the government wants to see more research into safety risks and potential health benefits of cannabis before legalizing it. The process is expected to take years.

But perhaps there’s some humanism in that. Selling a product — even one that’s standardized — before it’s been safety-tested isn’t something that even ChatGPT would suggest.