AI: Hungry for Power

Mike Hollan, Assistant Managing Editor, Pharmaceutical Executive

Artificial Intelligence (AI) continues to dominate the technology conversation in pharma and life sciences. While there’s still a lot of excitement surrounding the algorithmic tech, a worrying trend is starting to appear. The vast majority of conversations about AI are focused on what these tools could potentially do, or will eventually do one day. A lot of words have been written about how much work AI will eventually save us in the future.

The conversation is often focused on the future; there isn’t a whole lot of talk about what AI has actually accomplished.

That’s not to say, of course, that AI hasn’t been used to do impressive things. A common application of the technology is in drug development, where it’s used to find new patterns in the data, which can then be used to identify new molecules. While a number of drugs discovered or developed by AI have entered clinical trials, it’s still too early to tell how much more effective AI is at this process than traditional methods. In fact, some studies suggest that while AI-driven drugs might have higher success rates in Phase I trials, they appear to level off during Phase II and match standard industry success rates.¹

Granted, this data is based on a limited sample size, but it’s worth pointing out that implementing AI into R&D has yet to bring forth a massive flood of new molecules. Once again, the promise of AI lays in a potential future.

However, there’s another hurdle that the industry must consider before fully embracing AI: the energy costs. The technology behind AI consumes massive amounts of energy. According to reporting from Vox, a single ChatGPT query consumes nearly 10 times the amount of energy as a typical Google search.

This might be a personal opinion, but asking ChatGPT questions doesn’t appear to deliver results that are 10 times superior to traditional search engines. In fact, it often seems slightly inferior. Searching for information on Google might require a few more clicks to find the answer, but I’m much less likely to encounter hallucinations that way.

It’s impossible to discuss the potential of AI without discussing the cost, and the energy consumption component will absolutely impact this. Even if AI is able to produce results at a superior pace than traditional methods, those results must be measured against the energy consumption. For example, a car that drives twice as fast as the average car but consumes 10 times the amount of gas wouldn’t be considered an upgrade.

And, by today’s standards, we have yet to prove that AI can actually drive twice as fast.

It’s hard for the average user to comprehend the actual energy costs associated with AI because most users aren’t paying them right away. ChatGPT remains a free service, and Amazon has yet to raise its prices despite adding an AI assistant to its site. However, those costs will have to be paid, and not always in obvious ways.

The data centers that house the servers these systems run on consume a lot of water. As I write this column, massive wildfires are spreading across Southern California. Not only are firefighters struggling with a water shortage, one of the main reasons why these fires are able to spread so quickly is due to drought conditions. Massive data centers have been accused of hoarding water to cool the servers they house. This water hoarding has been linked to drought conditions across the world. AI didn’t spark the wildfires, but it did a play a role in making them more difficult to control.

It’s easy to become enamored with new technology. Over the past 30 years, the world has completely changed due to advancements, and no one wants to be left behind. If AI can truly improve the ways that pharma discovers new drugs, that’s important work and it would improve the lives of humanity.

On the other hand, it’s also important to remember that the hype around new technology isn’t always warranted. A lot of people spent a lot of money on NFTs that aren’t likely to become worth millions. Despite Mark Zuckerberg’s best attempts, we’re not all living in a VR-powered metaverse.

AI offers huge potential, but a lot of technologies have offered potential over the years. This isn’t meant to suggest that pharma should avoid AI. However, it’s probably a good idea to start considering the true cost of this technology before the industry decides to become completely dependent on it.

Mike Hollan is Pharm Exec’s Assistant Managing Editor. He can be reached at mhollan@mjhlifesciences.com.

Reference

1. Jayatunga, M.K.; Ayers, M.; Bruens, L.; Jayanth, D.; Meier, C. How Successful are AI-Discovered Drugs in Clinical Trials? A First Analysis and Emerging Lessons. Drug Discov Today. 2024. 29 (6). https://pubmed.ncbi.nlm.nih.gov/38692505/