Generative AI is the new darling of Silicon Valley. But what the heck is it? And what does it mean for the future of our jobs, yours and mine?
Experts believe generative AI will soon enter workplaces: Sequoia predicts that by 2023, generative AI will be able to put together scientific papers and visual design mock-ups; and by 2030, it will write, design, and code better than human professionals in the field.
Yet, few of us have a clear idea of how this will actually play out. How will it all begin? What will my job look like once it’s rolled out? It’s likely that your bosses also don’t know the exact answers to that, which is why it’s important to delve into the technology for what it is, and what it isn’t.
I have spent over 2,000 hours working with generative AI, and have built a company that creates AI bots to transform the future of work. I’ve come to believe generative AI is likely to be the biggest driver of productivity since the steam engine, but only if we understand both its potential and limitations.
For insurance, I believe that generative AI won’t kick every creative worker out of their jobs, but it will change how they go about them, and where their time and energy will be focused.
Here’s what generative AI can (and can’t) do, and how it will impact the way we work:
What is generative AI?
Generative AI is essentially a very very advanced form of predictive text. Generative AI allows users to insert text prompts and get back a piece of artwork, a blog post, or a sarcastic response to a question.
But how does it churn out that information? Has it become smart, taking your input and giving you something entirely new? Does it have an algorithm to respond to any worldly input?
Advanced AI models have digested hundreds of billions of words. Now, they can predict the most likely word and phrase combinations. This allows generative AI to suggest what word you may want to type next. But while you could ask generative AI to tell you a joke, it can only respond using the dataset it has processed. So while AI bots may seem like they understand instructions, it is not actually comparable to “understanding.” It’s more like an elaborate autocomplete.
For example, if you take a generative AI bot and give it the prompt 2+2=, it will respond with “2+2=4.” But that’s not because it has an internal algorithm, like a calculator, that has processed your request. It has just deduced from the entirety of the internet that the most likely answer to 2+2 is indeed 4. In this particular case, it’s also factually correct.
That said, a great autocomplete can be really productive. It can essentially take your unstructured ideas, notes, and drawings, and produce something beautiful. A rough brainstorm can become the first draft of an article. A sketch, the prototype of your next digital product. A list of strategic notes, an action plan. But while those outcomes can be terrific, they’re not the finished product, and shouldn’t be treated as such.
Will generative AI change the way I work?
In short, yes. But it may be limited by nature.
The first step to integrating AI into your workplace is to understand its limitations. After being fed billions of data points, AI has the theoretical intelligence of an adult, but the real-life judgment of a two year old. Meaning, it’s very good at following instructions, but terrible at knowing when or if they are the right thing to do.
Take for example a simple task: taking a list of bullet-points on a certain topic, and writing a blog post. Generative AI can do a great job at this. But it won’t know who the audience is or what buzzwords will make readers’ eyes roll. It won’t know what blog posts you’ve written before and what nuances triggered performance gains. And it won’t know when it’s time to do something totally new because what you’re doing now simply isn’t generating results. All it knows is what it learned from what others have written on the internet.
This weakness of lacking context goes even further. Although AI can look and sound human, it doesn’t actually know what world we live in. For instance, Generative Pre-trained Transformer 3 (known as GPT-3) is a generative AI model that uses deep learning to produce human-like text. But GPT-3 was trained on an index of the internet from 2016. Ask it who the US president is, it will tell you it’s Donald Trump. Ask it to use references to pop culture, and it will likely be outdated. It will execute tasks blindly, but could spit out responses that are simply incorrect.
When this type of misinformation looks authoritative and well put together, it has the potential to do massive damage inside large organizations where assets are constantly circulated with little context.
Because of that, generative AI today can only be trusted to take on very well-defined activities. And that, only with a robust, custom framework to guide it and review any content before it’s deployed. This isn’t to say that the technology isn’t game changing. But if you’re a CEO hoping that AI will replace the thinking of your best employee, that’s unlikely to happen any time soon.
How do I harness generative AI?
I believe that in the near term, AI won’t replace most jobs. But by taking on tasks that are not mentally demanding, yet highly time consuming, it could free up time in an employee’s day to do everything that AI can’t do—things that require high-level human insight, empathy, and critical thinking. Here are three examples:
1. Faster writing
Generative AI can speed up the process of writing, from articles to website copy. You can write down a few bullet points on your core message, run it through a program like copy.AI, and get a post that’s two-thirds of the way there in seconds. You may need a few rounds of reviews and edits, but you can still save time. That means you can spend more time delving deeper into stories, analyzing what topics are stirring interest, and meeting with people.
2. Improve customer service
Customer-facing roles also have multiple uses for generative AI. An employee can get a transcript of any conversation and get generative AI to produce an analytical summary of what was said. It can break down elements like: What’s the caller’s main pain point? What does the caller desire from you? AI can quickly filter out the unuseful details from a conversation.
3. Jump-start product mock-ups
Product designers can use generative AI to create basic visual mock-ups of ideas without spending hours in front of a computer. By building out the basic scaffolding at the early stages, before feedback and modifications, this technology can give workers more time for creative exploration with customers.
When you look at those three examples, what’s the common thread? Well, all these highly useful tasks still assume a human is architecting the work to be done. Because the AI still isn’t creating a single original thought. By contrast, what you bring to the table is a deeper relationship with the customer and translating that into clearly defined units of work that the AI can help execute.
That’s the true value of generative AI in the workplace: to remove time-consuming tasks that aren’t using employees’ brainpower, leaving you with time to get on with everything that is “unautomatable:” interacting with prospective customers, finding out what makes them tick, brainstorming innovative solutions for their individual needs, tweaking the product to meet their goals, learning from example.
I believe that every workplace needs to clear away the misconceptions of generative AI so we’re empowered by it, not using it irresponsibly by thinking it will replace those high-level tasks. It won’t replace humans. But it will revolutionize the future of work by giving people back invaluable time to do the work that really matters.
Erol Toker is the founder and CEO of Truly, a hyperautomation platform built for revenue teams.
Recognize your company's culture of innovation by applying to this year's Best Workplaces for Innovators Awards before the extended deadline, April 12.
