Chatbots have been around for years and years. As early as 1966, the first chatbot named Eliza was born, but it was only able to answer a few simple questions that had to be asked in a simple and straightforward way. With the advancement of technologies such as Natural Language Processing (NLP), Artificial Intelligence (AI), and more specifically technologies like GPT-3 that can be used to generate new text that is grammatically correct. In addition, chatbots can now answer pretty much any question without having the answers hardcoded by connecting to large answer repositories like Wikipedia. Some of the AI models from Google and Microsoft are so advanced in language understanding that some benchmarks show that they are exceeding humans. Interestingly enough, they both exceeded the human baseline score of 89.8. The score is a percentage that is calculated for each of the tasks based on their individual metrics.
Beyond language understanding, chatbots now can also translate text, answer questions, summarize texts, write essays, write code, and even draw images based on a description. The new, soon to be released, AI from OpenAI, it’s called Dall-e (a combination of artist Salvador Dalí and Pixar’s WALL•E). It’s a chatbot that allows you to type something, and it returns images interpreting what you are saying; if you type, “show me a chair with the shape of an avocado,” it could return something like this:
The same way that GPT-3 can write endless pieces of text, Dall-e can create endless images that can be either photo-realistic or look like drawings.
Beyond language understanding, chatbots now can also translate text, answer questions, summarize texts, write essays, write code, and even draw images based on a description
The COVID-19 pandemic has also kept everyone at home, providing additional interest in chatbots. Chatbots are a great way to interact with brands, they are used for customer service and more and more on social platforms. The first venue to support chatbots was Facebook Messenger, but shortly after, many other platforms followed suit: Telegram, Skype, Viber, Slack, Microsoft Teams, Twitter, Google Assistant, WeChat, Line, WhatsApp, and many more.
With the combination of these and other technologies, today’s chatbots are pretty smart and powerful (definitely smarter than Eliza, the chatbot of 1966!), but what will the chatbots of the future look like? Will they be a fully immersive experience? Will you chat in virtual reality? Will you have a hologram customer service representative?
I think the chatbots of the future will look and sound more human. These bots should be able to respond to emotions as well as participate in actual conversations. More importantly, the chatbots of the future will take on a larger share of the cognitive load. Instead of using a search engine, you could ask a chatbot what you want to know, and it will find the most relevant information for you. Imagine interacting with a chatbot over video chat (or hologram chat) and this celebrity or expert in the field, that looks and sounds human, is a bot that can have a conversation with you and answer your questions by sorting and filtering information. The two technologies that can power cognitive capabilities (in addition to the ones I mentioned before) are voice cloning and deep fakes; there are many companies doing this today. For example, for voice cloning, I was recently playing around with resemble.ai and 15.ai. Resemble.ai lets you record a few messages to build your voice. From there, you can type any text and it will read it with your voice. 15.ai has a preset of characters (like Spongebob or Dr. Who) to which you can select and have them say anything you type. There are also some companies working on Deep Fakes, like synthesia.io, which lets you generate deep fake videos. We used it for a campaign called Malaria must die to recreate a video of David Beckham in many different languages.
While this future is pretty impressive, all these technologies can be used to impersonate or scam others. Because of this, more companies are also creating tools to detect fake voices, deep fake videos, and AI-generated text. Microsoft has a tool called “Microsoft Video Authenticator”, and it can analyze video or a still photo and provide a percentage chance, or confidence score that the media is artificially manipulated. Microsoft also created a quiz to help people identify deep fakes. Another tool called “Reality Defender” helps you detect synthetic media, and for audio, Resemble.ai created an open-source tool to help you detect fake speeches. There are also tools to detect generated text like https://gltr.io/ and “GPTrue or False”, a chrome extension that helps you detect if the text was written by GPT or not.
What do you think the chatbot of the future looks like?