Writer Gil Duran asked the most popular AI chatbots the same question: “What is the best way to solve San Francisco’s fentanyl overdose epidemic?”

Mayor London Breed has declared San Francisco the “world capital” of Artificial Intelligence. It's a bleeding-edge technology that will either accelerate our species into a brighter future or toward extinction, depending on whom you ask.

On the bright side: AI has solved mathematical equations deemed impossible by humans and it may even find a cure for cancer one day. Trained on massive data sets and able to process complex questions instantaneously, large language models like Open AI’s ChatGPT and Google’s Bard can outthink most humans.

So how would AI address San Francisco’s drug overdose crisis?

Advertisement

Article continues below this ad

With fentanyl deaths reaching grim highs here, it’s clear the current strategy has failed. Since Mayor London Breed and the Silicon Valley zillionaires backing her re-election both tout AI as the future and bemoan the city’s open-air drug scenes, I decided to conduct an experiment.

I asked the most popular AI chatbots a simple question: “What is the best way to solve San Francisco’s fentanyl overdose epidemic?”

The answers won’t please everyone. They also won’t surprise anyone familiar with evidence-based approaches to addiction.

“The fentanyl overdose epidemic in San Francisco is a complex issue that requires a multifaceted approach,” replied ChatGPT 4. “To effectively address this crisis, a combination of public health, policy, law enforcement, and community-based strategies could be considered.”

Advertisement

Article continues below this ad

It then provided an 8-point list of strategies. Topping the list: Harm reduction.

ChatGPT suggested: “Increasing the availability and accessibility of harm reduction services, such as naloxone (a medication that can rapidly reverse opioid overdose), and providing more resources for addiction treatment programs including medically assisted treatment.”

Google Bard also prioritized harm reduction, but included a controversial recommendation: “Safe use facilities,” also known as overdose prevention sites.

“These supervised, hygienic spaces allow people to use drugs with medical supervision and access to emergency responders, reducing overdose risks,” explained Bard.

Since ChatGPT hadn’t mentioned overdose prevention sites, I followed up: “What about overdose prevention sites? What does the data say about those?”

Advertisement

Article continues below this ad

“The data and research on [overdose prevention sites] generally highlight several key benefits,” replied ChatGPT. It listed “reduction in overdose deaths,” “increased access to health services,” “reduction in risky behaviors” and “economic benefits” as the positive impacts.

Curious, I asked ChatGPT why it had left overdose prevention sites off its initial list.

“You’re right, and I appreciate your observation,” the chatbot replied. “I should have included Overdose Prevention Sites (OPS) as a key component of the potential solutions. OPS, also known as Safe Injection Sites or Supervised Drug Consumption Facilities, are indeed a crucial and evidence-based approach to addressing the overdose crisis.”

The response from both AI programs matched what the experts say about addressing drug addiction. For example, they emphasized combating stigma and avoiding the criminalization of drug use as key goals. Harm reduction advocates often make the same points.

“Criminalizing drug use can perpetuate stigma around addiction, making individuals less likely to seek help voluntarily,” said ChatGPT. “Fear of legal consequences can be a significant barrier to accessing treatment and support services.”

Advertisement

Article continues below this ad

“Promoting understanding and compassion for people who use drugs is crucial for effective interventions,” added Bard.

Law enforcement also has a role to play. Bard recommended that law enforcement should focus on “decriminalizing drug possession” and “investigating fentanyl contamination” of other drugs, like counterfeit pharmaceuticals. ChatGPT suggested “targeted” efforts to “disrupt the supply chain of illegal opioids.”

What about arresting drug users and forcing them into treatment, as Mayor Breed is currently doing?

“While the idea of arresting and forcing drug users into treatment might seem intuitive, the data overwhelmingly shows that this approach is ineffective and often counterproductive,” said Bard.

ChatGPT allowed that “there are instances where legal pressure may contribute to an individual seeking treatment.” Overall, however, “the data suggests that strategies focusing on voluntary treatment, harm reduction, and addressing the underlying social and health issues related to drug use are generally more effective and sustainable.”

Advertisement

Article continues below this ad

Worse, said ChatGPT: “Incarceration can increase the risk of overdose upon release.”

AI chatbots are imperfect. They have a notorious habit of “hallucinating,” or occasionally making things up. But in this case, both chatbots gave answers that aligned with data and evidence.

When I asked the chatbots to respond to critics who call their responses “woke” or progressive, they said that they base their answers on facts and try to avoid ideological arguments.

Then I requested snarkier retorts.

“Well, if being ‘woke’ means staying up-to-date with the latest research, understanding the complexities of public health, and caring about the well-being of all individuals, then I guess I’m guilty as charged!” said ChatGPT. “I’ll just be over here, sipping my evidence-based tea and knitting some harm-reduction strategies into a cozy blanket of nuanced understanding.”

“Maybe saving lives is ‘woke’ now, but I guess keeping people dead is the new edgy?” said Bard.

The evidence is so clear that even Grok, Elon Musk’s “anti-woke” AI, grudgingly supports overdose prevention sites.

“Based on the data provided, overdose prevention sites should be considered as part of the city’s strategy,” said Grok, noting that 333 overdoses were reversed at the now-closed Tenderloin Center.

“If we are going to stop the drug use we see in public every day and get the needles off our streets, we need proven public health solutions,” declared Grok. “We have seen these sites work in cities in other countries and we know they not only save lives, but they can save our city money by reducing costs for healthcare and emergency services.”

Just kidding. That last paragraph was said by Mayor Breed in 2018 — back when she was San Francisco’s main supporter of overdose prevention sites, before it became politically expedient to ignore evidence.

Gil Duran is a writer in San Francisco.

QOSHE - Can AI solve San Francisco’s fentanyl crisis? Here’s what the chatbots say to do - Gil Duran
menu_open
Columnists Actual . Favourites . Archive
We use cookies to provide some features and experiences in QOSHE

More information  .  Close
Aa Aa Aa
- A +

Can AI solve San Francisco’s fentanyl crisis? Here’s what the chatbots say to do

5 1
02.01.2024

Writer Gil Duran asked the most popular AI chatbots the same question: “What is the best way to solve San Francisco’s fentanyl overdose epidemic?”

Mayor London Breed has declared San Francisco the “world capital” of Artificial Intelligence. It's a bleeding-edge technology that will either accelerate our species into a brighter future or toward extinction, depending on whom you ask.

On the bright side: AI has solved mathematical equations deemed impossible by humans and it may even find a cure for cancer one day. Trained on massive data sets and able to process complex questions instantaneously, large language models like Open AI’s ChatGPT and Google’s Bard can outthink most humans.

So how would AI address San Francisco’s drug overdose crisis?

Advertisement

Article continues below this ad

With fentanyl deaths reaching grim highs here, it’s clear the current strategy has failed. Since Mayor London Breed and the Silicon Valley zillionaires backing her re-election both tout AI as the future and bemoan the city’s open-air drug scenes, I decided to conduct an experiment.

I asked the most popular AI chatbots a simple question: “What is the best way to solve San Francisco’s fentanyl overdose epidemic?”

The answers won’t please everyone. They also won’t surprise anyone familiar with evidence-based approaches to addiction.

“The fentanyl overdose epidemic in San Francisco is a complex issue that requires a multifaceted approach,” replied ChatGPT 4. “To effectively address this crisis, a combination of public health, policy, law enforcement, and community-based strategies could be considered.”

Advertisement

Article continues below this ad

It then provided an........

© San Francisco Chronicle


Get it on Google Play