It is very clear that a large number of people did not expect the viral acceptance of ChatGPT, not even OpenAI was able to predict this craze. Before it made history as a consumer tool at the fastest growth rate, before the term “pre-trained generative adapters” was popularized, and before every company you could imagine rushed to implement their core model, ChatGPT debuted in November as a “research preview.”
The first blog post describing ChatGPT as just a simple fun project has since become an irrelevant reference because it probably undermines the very essence of what it has become today. The post introduced “ChatGPT as a sister model to InstructGPT, which is trained to follow instructions in a prompt and provide a detailed response. We’re excited to introduce ChatGPT to get user feedback and learn about its strengths and weaknesses.” There is no poetic verse about how technology is profoundly changing the way we interact with it, not even a syllable about how amazing it is. Just a research preview, really.
Just four months after its latest innovation, ChatGPT is actively changing the way the world looks at technology. Because of the direction things are headed, metaverse or glitzy interfaces are not the technology of the future. “Type commands into a text box on your computer” is what it is. The command line is back, but it just got smarter this time.
In fact, generative AI is moving in two directions at once. First, which adds new tools and capabilities to the things you’re already using, is more focused on infrastructure. Massive language models such as Google’s GPT-4 and LaMDA will help you write emails and notes, automatically optimize presentation suites, fix any errors in your spreadsheets, edit your images more efficiently than you can, and in many cases, create Your code for you.
In general, this has been the path of artificial intelligence for a while, right? Over the past few years, Google has integrated various types of AI into its products, and even companies like Salesforce have developed aggressive AI research initiatives. These models can revolutionize corporate efficiency, but they are expensive to develop, expensive to train, and expensive to query. AI improvements in the goods you use now is and will continue to be a major industry, or at least receive significant investment.
The second direction of AI, where interaction with AI becomes a consumer product, has been much less clear. Now thinking about it again, it makes a lot more sense, who wouldn’t want to interact with a robot knowledgeable about movies, cuisine, things to do in Tokyo, and more? Who would have thought that typing in a chat window would become the newest big thing in user interfaces before ChatGPT took the entire planet by storm and before Bing and Bard tried to build their own products on it? This is, in a sense, a return to a very old concept. For a long time, majority of users used computers only to type commands in blank screens using command line.
But then, we’ve created better user interfaces! The problem with the command line was that you had to be very precise about what you typed and in what order to get the computer to do what you wanted. It was much easier to point and click on the large icons, and it was much easier to explain to people what a computer could accomplish with pictures and icons. The GUI still dominates even though it’s not the most complex interface, messaging may be the most expendable. Consider Slack: You generally think of it as a messaging app, but you can integrate links, editable documents, interactive polls, educational bots, and more into that back-and-forth interface. WeChat is famous for being an entire platform — basically, the entire internet — condensed into a messaging app. Lets you start with a message to navigate away from the command line it replaced.
However, the developers never gave up trying to create a chat UI functionality. An excellent example is WhatsApp, which has spent years trying to understand how customers can use chat to communicate with businesses. Allo, one of Google’s many unsuccessful messaging apps, hopes that you can talk to an AI assistant while chatting with your friends. Many smart people thought messaging apps were the future of everything during the initial chatbot mania, which peaked around 2016.
The messaging interface, or “conversational AI,” has a certain appeal. It all starts with the fact that we all know how to use messaging apps, on which we spend a lot of time and effort because it is how we keep in touch with the people we care about the most. While you may not know how to find out your frequent flyer number on the Southwest app or how to traverse the Uber app’s hidden areas, practically everyone can understand the concept of “send these words to this number.” Messaging can greatly simplify experiences in a market where users don’t want to download apps and mobile websites are still generally poor.
But many of these apps face the same issues. Chat is ideal for short exchanges of information, such as during business hours; You can ask a question and get a quick response. But are you searching through an index in the form of messages? never. Buying an airline ticket after exchanging 1000 messages? No thank you. It’s the same voice assistant, and if you try to use Alexa even to make a basic purchase. If you ask me, the dedicated visual UI is quite effective than the messaging window for most complex tasks.
And things quickly get tricky when talking about ChatGPT, Bard, Bing, and the rest. These models are smart and cooperative, but in order to get what you want, you still need to know exactly what to order, how to order, and in what order. The concept of an “instant engineer”, someone you pay for the specialized knowledge needed to get the perfect image from Stable Diffusion or to have ChatGPT write the perfect JavaScript, might sound silly, but it’s an absolutely essential component of the solution. It is similar to the early days of computing when only a few were able to instruct a machine on what to do. There are actually websites where you can buy and trade really good claims, expert mentors, and books about claims. I also imagine that Stanford is already working on an accelerated engineering major that will be available to everyone in the near future.
Generative AI is amazing because it seems to be capable of practically anything. This is also the root of the issue. What do you do when you have unlimited options? How do you get started? When all you have to look at to see what it’s capable of is a blinking indicator, how can you learn to use it? These companies may eventually provide more interactive visual tools that help people understand what they can accomplish and how it all works. This is one reason to pay attention to ChatGPT’s new plug-in system, which is currently fairly simple but may soon expand the functionality of the chat window. The greatest idea any of them have now is to make some recommendations for what you might write.
AI was intended to be a feature of AI, but it has now managed to turn into a product. As a result, the text box returns. Messaging has become the interface again.