Artificial Intelligence isn’t new. The term was first used in 1956 at Dartmouth College. And it has gone through its phases ever since. These phases are termed either summer, where a lot of interest and funding happens. And winters, in which the interest wanes and funding is turned down. Winters tend to happen mostly (I believe) because of disillusion, the fact that the AI didn’t end up delivering that end game (not the Marvel movie) type of technology.
Previous AI winters
The term winter relates to a cooling of interest and funding regarding the topic. There is a great article on AI winters in particular on Wikipedia that you can find here.
Previously and currently, people might have over promised Gen AI’s potential. I do believe we are looking at a very powerful tool, that can increase productivity of the workforce a whole lot.
But in the end, I think that many people will attribute more capabilities to the tool than it can muster. We talked about this previously, but it’s good to remember that Generative AI is primarily a prediction engine, that finds the best probable answer to your question word by word, without knowing much more about the context.
As an example, I recently took Copilot in Planner for a spin. It was only able to perform 7 out of 19 prompts. Most of these failing prompts were typical Project Manager actions, such as adding a duration to a task. Or adding dependencies to tasks. Even the more modern feature of goals, and adding tasks to these goals was a bridge too far for the AI to handle.
Copilot in Planner is currently in preview, but still, it was likely build using the latest AI models. I believe, the Copilot wasn’t trained enough on basic project managers needs. It’s capable of building a task list, and adding or removing tasks from the list. But not much more.
And that will likely be the disillusion that throws Gen AI into the freezer: the fact that we still need to train or prime a generative AI to our specific needs. Which is something that most people probably aren’t excited enough about to do.
The current spike in generative AI interest
Billions have been poured into the topic of Generative AI. OpenAI (creators of ChatGPT) showed up almost out of the blue and took the world by storm.
As we know, Microsoft is investing a lot in getting Generative AI into (almost) every application they have. Stating that it’s the next big thing since the Graphical User Interface (or GUI for short).
And it’s save to say, that the current spike is a very interesting one. One, that even got the interest of people that aren’t PC wizards. I for one changed the focus of the blog completely and contribute most of the writing to help the PMO learn more about AI and how it can help them in their work.
Generative AI, is closer to the end user than other Artificial Intelligence applications. There by making a lot of work easier to perform, like drafting up mails or complete novels.
But that’s where the issue lies I believe, they are still drafts.
Why do I think a winter is coming?
Well, it might just be the time of year (August 2024) and the warmer days are starting to be a thing of the past again. But it might also be this dip in the Google search trends:
But it could also be on the horizon due to some recent stock market crashes. The causes might not have been AI’s lack of delivering, but if there’s less money in the market, there will be less oppertunity to invest in getting AI “there”.
The issue with Generative AI drafts
I mentioned this above, the issue might just be that AI isn’t fully capable of understanding our intentions or our creative thought process. Drafted e-mails tend to look stale, too formal, or they just don’t deliver the message the way we would.
And that’s e-mail. A text heavy interface. Other applications suffer even more. With PowerPoint for instance, it’s not capable of delivering a presentation in the style I’d like it to have. Even when I present it with the themed template and a clear prompt to keep in line with previous presentations I’ve done.
When you need to write something short and quick, Gen AI can get you to 80%. When writing manuals or guides it might even get you close to that 80%.
But you, as an end user, still need to push it over the finish line.
And that last bit, might be the reason why people start to lose interest. It’s a great tool… But it’s still “just” a tool. And not a magic wand, like so many people have mentioned it would be.
Can we prevent a Generative AI winter?
Maybe it’s only human nature to lose interest in a topic after a while. Maybe we get “AI attention fatigue”. Which doesn’t mean that we are ungrateful or don’t use the tool. It might just mean that we are starting to consider it “just a part of the normal workday”.
As long as it’s not capable of delivering that new Wow-moment. The moment where it hits the mark every time we prompt it to do something specific (like adding durations to a Gantt task 😊). The masses will lose interest in Gen AI.
It might be capable of taking 80% of the work away, it might even be close to 100% when it comes to coding. But a human is still required to finish the work, thankfully!
Apart from its capability, there is also the funding topic. If markets crash (for what ever reason) and the money stops pouring in, AI might just stop improving. Combined with that waning interest in the topic, people might fund the research less.
That said, there’s a lot of money already in AI, it might just be enough to ride the wave. Maybe there’s “AI global warming”, and we will never cool our interest.
Generative AI is still a very new field within the general AI topic. So, maybe I’m completely wrong, and I’d be the first person to admit it. I just wanted to write down some thoughts, I hope you enjoyed reading them.
Final notes
I think the disillusion of Gen AI is a good thing. We can use it to keep a solid foot on the ground and continue with more important things, like doing our jobs, but way more efficient.
Maybe Gen AI will go “the way of the GUI”, I don’t think about the graphical interface of the applications I use. But still, I wouldn’t be able to work as efficient with those tools if there wasn’t a graphical interface. Maybe in the future, we won’t know how to work without Generative AI, but we won’t care about it very much either.
I’ll keep doing these (weekly) blogposts on Gen AI and the PMO for now. But I believe there will be less interest in the near future regarding the topic.