People are excited because it’s a clear jump forward from a technological standpoint. Ryan Rascop, a senior business development manager of AI at Insight, said as much during a recent Insight webinar titled “ChatGPT: AI Ethics & Business Impact.” He pointed to the tool’s versatility, i.e., how it creates content and answers questions with a high degree of accuracy without it simply regurgitating them from a website as differentiating factors.
“It’s as close as we’ve gotten or seen in a very media-forward position of a chatbot… mimicking human responses to questions,” he said. “To me, this is the dawn of a new age in IT and what we’re able to do, where we’re able to gather information and how it will change our daily lives.”
Regarding the “high degree of accuracy,” it has undeniably made mistakes. ChatGPT does of course have limitations, but that’s true of just about anything and, if you look at how far we’ve come in the world of AI, it’s very far indeed, virtually light years ahead of the earliest chatbots (which have advanced so much that “conversational AI” is probably more of an accurate term these days).
For example, Dr. Sbaitso (Sound Blaster Acting Intelligent Text to Speech Operator) from Creative Labs was fairly advanced for its time, over 30 years ago. Framed as a game for all intents and purposes, it saw the “doctor” play the role of a psychologist, asking the user to confide in them their problems.
It was more than anything else a showcase of the capabilities of the company’s Sound Blaster sound-card technology. However, Dr. Sbaitso could nevertheless carry on a conversation with the user, albeit one scripted on the doctor’s end of things. Fast-forward to today, and scripts are out the window altogether with ChatGPT, even if a given user is happy to simply text back and forth with it in similar fashion. As impressive as that may be in and of itself, it’s also just scratching the surface of the AI’s potential, though.
The possible functions range from the entertaining (having elaborate conversations with an AI) to the ethically dubious (passing off AI-composed essays as one’s own work in school — hence the curious preamble at the top of this piece). However, in the middle of that wide range somewhere also sits a subset of increasingly practical use cases.
It shouldn’t exactly come as a surprise. Hollywood and science-fiction literature is filled with examples of AI taking the next steps in its development (with varying degrees of success as far as the fate of humanity is concerned). However, if you can look past the hypothetical future made famous by The Matrix (for example), one that’s far more utopian in nature takes focus, one that makes all our lives easier.
It goes beyond the everyday tasks already touched on too. Think applications in the business world, like creating code. However, in what was a recurring theme during the aforementioned Insight webinar, the idea is more so to eliminate tedious tasks so you can focus on what’s most important.
“I don’t think it’s about replacing humans. I think it’s about augmenting our skill sets. So, there may be a component of tech jobs in the future where you will need to utilize these types of tools,” Rascop said, using an example of potentially using ChatGPT to create a Microsoft PowerPoint presentation. You’re theoretically inputting the data, asking it to work to its strengths putting together a slideshow. You still need to know the topic and present it.
It's worth noting, when prompting any content AI, whether for prose, like “write me an essay about wolves,” programming code (“show me how to implement a search function)” or for visual content (“paint a futuristic Eiffel Tower in the style of van Gogh) the output depends on the quality of the model, but also the suitability of the prompt. You have to consider things like the volume and diversity of inputs contributing to the model, flexibility provided by the model for in-prompt tuning and the level of specificity of the prompt itself. Look for a new technical discipline of “promptcraft” or “prompt engineering” to emerge, joining existing technical disciplines to round out the tool kits of information workers.
Remember, our own mind is still the best and most sophisticated tool we have, as we further explore how these new modes of interacting with large bodies of data will impact our daily lives and our productivity. The best strategy is to accelerate our own ability to be productive and creative by leveraging AI and other emerging technology.
What we at Insight are specifically on the lookout for from a transformation perspective, is Microsoft, a widely publicized investor in OpenAI, integrating AI functionality with its Azure portfolio. We’re talking at least to a greater extent than Microsoft has to this point anyway. Microsoft has already granted Azure customers access to advanced AI models, enabling advancements in areas like data extraction and analysis and customer support. In that last regard, AI can for example summarize tickets to free up support employees’ time.
We’re also anticipating productivity enhancements in programs like Microsoft PowerPoint, as described above, or Word — think a “Clippy-style” office assistant updated for the modern workplace. After all, in the here and now, you can ask ChatGPT for step-by-step instructions on pretty much whatever you can think of, including how to use Microsoft 365 software.
You can literally ask the chatbot for the right Microsoft Excel formula — then how to implement it in your spreadsheet. There comes a point when you have to ask yourself is this how the next search engine will work, complete with the ability to refine searches with follow-up questions, as if you were chatting with a real-life person. The answer is an irrefutable yes.
OpenAI is of course the creator of ChatGPT — terms like “creator” taking on new meaning in this context. However, it’s Microsoft that has integrated ChatGPT in a new version of its Bing search engine, now currently in limited preview mode. So, it’s realistically a matter of time before searching the internet means less time spent sifting through pages of results and ends up having more in common with a pleasant conversation with a subject matter expert, one who has gone to the trouble of referencing sources in its responses, no less.
That having been said, it’s far from simply academic at this juncture. There are undeniable kinks that need to be worked out, as the chatbot has to be reined in some. For example, in one instance, it declared its love for a New York Times columnist, asking him to leave his wife. In another instance, it revealed confidential information, all prompting Microsoft to implement a chat limit for users (instead of unlimited chats), as the AI had not been tested internally to carry out long conversations. That’s arguably fair, though.
After all, what’s the point of a limited preview if not to test for limitations and make adjustments? To its credit, even if the Bing chatbot isn’t ready yet, Microsoft has undeniably addressed some of ChatGPT’s shortcomings. According to OpenAI, ChatGPT is not connected to the internet and is dependent on its training data from 2021, meaning it’s not up to speed on what’s happened over the last few years. So, you can’t really ask it any questions pertaining to anything going on in the world today expecting an accurate response (other than it doesn’t have information “from the future”). In comparison, the Bing chatbot has internet access and can correctly answer questions on current events with relative ease.
Now, Bing’s not necessarily the answer here. It’s at least a work in progress, but there’s good reason to believe, whatever its issues, they will get resolved eventually. They’ll have to be, considering all the competition Microsoft is facing. For example, Google is setting up its Bard AI chatbot as a direct competitor. With more general regard to ChatGPT, alternative AI chatbots/writers like YouChat, Jasper and Notion AI are all throwing hats into the ring.
ChatGPT may indeed be the fastest growing app in internet history with 100 million users in two months, according to UBS. It’s also projected to earn $1 billion in revenue by next year, once it begins to monetize its services, which are generally free for right now (because it’s still in the testing phase). However, there are no guarantees it stays on top. A lot can change. What likely won’t though is society’s overwhelming appetite for innovation, growth, and change.
It’s more than just en vogue, as the technology has literally been building up for decades, much like chatbots in general. AI-powered chatbots may be relatively new, but they too have already been here for a few years. They’ve only been getting more sophisticated over time. Conversational AI’s next steps past this point are undeniably cause for justified speculation. ChatGPT probably has an interesting take on the matter. Maybe just ask it what it thinks.