Generative AI: What is it good for?
If you are insane college football fans like us, you’ve noticed a growing trend the past couple of years when watching football broadcasts. On 4th downs when a team must decide to go for the first down or punt, inevitably the broadcast shows the “what do the stats say to do” graphic. The same goes for the decision after a touchdown on whether to go for 2 points or to simply kick the extra point. The announcers discuss the same “what do the stats say to do” metric. Increasingly, many football teams are turning to data and algorithms to help them decide what decisions to make under certain conditions.
Of course, in any competitive environment like sports or business, decision-makers will turn to whatever advantages they can find. But the interesting thing about some of these football decisions is that they seem to produce some odd decisions. Some of those decisions turn out to be good ones, and some turn out to be just plain head-scratchers.
Our culture stands at a technological tipping point. AI and automation truly have (and will continue to) revolutionize all aspects of our society, but particularly in business. Elsewhere in this blog, we have written about several effective uses of AI for business owners.
However, the above head-scratching decisions in a highly competitive environment like football are instructive to business owners, as well. These should help you consider the advantages and limitations of AI in your business.
As you know, present generation AI works best in business at repetitive tasks with clear logic and boundaries – things like aggregating data and “entry level” customer service like recommendations or common questions. It also can be useful at gleaning content highlights (which is where those meeting helper softwares like Otter and Fireflies come into play). Some healthcare-oriented AI engines are increasingly accurate at diagnosing diseases based on inputs like symptoms, test results, and imaging. Again, there are plenty of AI applications that can likely help you depending on your industry and circumstances.
But one particular area folks often use AI for requires a bit more consideration – that is, what is called generative AI. This would include things like using AI models for creating (“generating”) blog posts, videos, or images for your business to use. Simply put, generative AI usage requires, at minimum, human effort in revision and editing.
Two Potential Limitations of Generative AI
Synthesis and Research
One notable tendency in generative AI is its inconsistency in identifying appropriate research AND in “synthesizing” research into claims. For example, AI engines have been known to completely fabricate sources whole cloth. These engines don’t ALWAYS do this, and it has become less common than when popular generators first arrived on the scene a few years back. However, this can still be a problem, and any use of generative AI for research requires thorough consideration and editorial oversight to ensure that the sources cited are, in fact, real. In fact, bad actors can even purposefully manipulate AI engines to create bad source material like deepfakes.
Additionally, generative AI can occasionally illustrate problems in synthesizing multiple sources of information. The ability to synthesize information – that is, to gather outside information from multiple sources, analyze it, and incorporate it appropriately into one’s own thinking – is a sign of high-level literacy and logic. Synthesis should be what separates a graduate student from a freshman. Synthesis should be what separates a master technician from an apprentice – the ability to bring multiple experiences and sources of information to a situation and apply it appropriately. AI is increasingly effective at this task. However, generative AI can often still struggle with this, showing an inability to “think” deeply and broadly about a topic.
Compounding the problem is how AI engines learn new data. Most generative AI engines are not “continuously learning” — that is, they don’t go out and search the entire internet for all the latest news and updates, and they don’t ingest new material given to it by users into their learning, either. This is a good thing overall so that engines aren’t swayed by unverified information. But it also means that AI engines may not register legitimate “latest and greatest” updates and information. Things that YOU would know are true from context may not be within the engine’s learned data.
If using generative AI, it is important that all facts and research are double-checked, but it’s also important to revise the material to include your own robust analysis and expertise. Generative AI engines (particularly text-based models) simply have not reached the ability to produce flawless, well-developed works.
It Follows the Rules, but it’s Average and Lacks Appeal
Similarly, generative AI does an excellent job in following rules. So, for example, text produced by AI engines is typically very clean. This is why programs like Grammarly and ChatGPT are useful to business owners looking to produce copy for websites, newsletters, and blogs. Grammar rules are clear and consistent, so this area is a good fit for AI.
However, besides the problem of research and source material mentioned above, generative AI also might be problematic because, well, it all sounds the same. It produces nothing unique. You should be paying attention to the blogs and web materials you are reading, as too many now are produced by bots whose trustworthiness extends only so far as what data set they are examining. In fact, a few years ago, the formerly renowned magazine Sports Illustrated was busted using bots and AI-generated content (without informing readers). This may be completely fine in your circumstances, but it’s also possible your audience may have different expectations.
If you want to establish trust with your audience, you’ll want to bring your own style to the content. AI engines CAN change style with some direction (try telling ChatGPT to write the same article in “bro” style, then in “pirate” style, and then in “Charles Dickens” style to see what it can do…the results are pretty funny). Again, it requires some human editorial effort to ensure some personability.
Back to those “Go for it on 4th down” decisions. AI is a game-changing tool, and successful businesses will consider how it can be used well within their industry and specific situation in order to gain a competitive advantage. However, it does have limitations, and at minimum human oversight and analysis are needed as we move forward into the brave new AI world.