Unnervingly lifelike, chatty, worldly-wise… you may mistake it to be the voice of the grandma living next door. Let’s call her ChatGPT, a free chat bot that can answer almost all questions with eloquence. All these words — ‘free’, ‘chat bot’ and ‘eloquence’ — have turned ChatGPT into a hobby for millions.
All you need to do is ask questions at a text prompt — ‘My parents are ageing and how do I make them see reason?’, ‘I’m trying to come up with some interesting, fantastical ways of decorating a living room for a design project’, ‘What does sunny side up mean?’… these are just a few questions that ChatGPT has been asked to answer by different people. We asked the first question only to be told: ‘Use reason and logic, show respect, seek common ground, seek outside help and be patient.’ All this was explained to us in 400-odd (sensible) words.
The language model has taken conversations around what AI can do deeper than ever before. It’s obvious that the chat bot’s capabilities are extraordinary as well as fascinating.
Inception moment
Behind the technology is the same organisation that launched the text-to-image generator Dall-E to the public. San Francisco-based AI company OpenAI has funding from investors like Microsoft, Reid Hoffman’s charitable foundation and Khosla Ventures (the brainchild of Vinod Khosla, the co-founder of Sun Microsystems). The face of the company is co-founder Sam Altman. (Elon Musk is a co-founder of OpenAI but he left the board in 2018.)
Like OpenAI’s other tools, ChatGPT (“generative pre-trained transformer”) has made a splash with people signing up in large numbers. My personal Twitter feed is packed with ChatGPT answers, many of which verge on sorcery.
Before its arrival, chat bots have been meh with some good answers coming now and then. That has completely changed. If you are not careful, the next marketing copy may come courtesy ChatGPT. The technology that’s being used is based on what the company calls “GPT-3.5”, an upgraded version of GPT-3, the AI text generator, which came out in 2020.
ChatGPT at work. Picture: The Telegraph
Challenge for Google?
You could be wondering what Google thinks of it. If you have noticed, Google answers in search are becoming more and more naturally phrased. At first, it was just about showing a few links to a query. That has evolved. Google already has BERT (Bidirectional Encoder Representations from Transformers), one of the first large language models developed by Google.
The company’s work on a language/conversation model called LaMDA (Language Model for Dialogue Applications) has already been showcased. It was unveiled at I/O 2021 “to converse on any topic” but Google back then said it was still in the R&D phase. Google also has MUM (Multitask Unified Model) that “allows people to naturally ask questions across different types of information”. And there is PaLM (Pathways Language Model). If MUM allows questions to be asked, LaMDA can hold conversations while PaLM is what can answer questions.
Despite having several language models in its arsenal, Google may not be comfortable about ChatGPT. During the recent Google for India conference, on being asked about the new disruptive chat bot, Google representatives more or less brushed aside the question. But a recent New York Times report says ChatGPT has led Google’s management to declare a “code red”. For over two decades, Google has been the world’s most known gateway to the Internet but competition is at hand for traditional search engines.
For Google, the answer to the new chat bot could be LaMDA, which received enough attention earlier this year when a (now fired) Google engineer, Blake Lemoine, claimed it was sentient. It wasn’t true but the technology appears to have improved in recent months.
ChatGPT doesn’t have a personality, instead it can blend knowledge collected from many sources with training from humans. Yes, these chat bots learn their skills by analysing huge amounts of data posted to the Internet. According to reports, unlike Google, ChatGPT doesn’t crawl the web for information on current events, and its knowledge is restricted to things it learned before 2021.
How it can be (and is being) used
OpenAI said the new AI was created with a focus on ease of use. “The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests,” OpenAI said in a post announcing the release.
So far we have seen ChatGPT being asked to write essays, college applications and even short scripts. The results are largely winding answers, with few specifics. Essays don’t evoke emotions but there’s clearly a voice in it, however mechanical. At times the output may appear vague but, at the same time, creepy and points to a nightmarish future. There is always that potential where chat bots can be used by unwanted political entities or make conversations around race and ethnicity scary.
Yet, what remains to be seen is how the likes of Alexa and Siri respond to ChatGPT. If the new technology becomes available as an application programming interface (or API), it will allow websites to create virtual assistants, customer service bots or marketing tools.
Generalisations about ChatGPT are in plenty, like will it ever replace writers. Computer scientist Yejin Choi, a 2022 recipient of the prestigious MacArthur “genius” grant, recently told The New York Times: “I don’t think it can replace anybody there because it doesn’t have a true understanding about the political backdrop and so cannot really write something relevant for readers…. People believe in tarot cards. People believe in conspiracy theories. So, of course, there will be people who believe in AI being sentient.”
Yet, the year 2022 will be remembered for the birth of ChatGPT. Here onwards, let’s fall back on Arthur C. Clarke’s words: “Any sufficiently advanced technology is indistinguishable from magic.”
At a glance: ChatGPT
What is it: ChatGPT is a language model chat bot developed by OpenAI based on GPT-3.5. It can interact in conversational dialogue form.
Can you use it: Yes, you can sign up and it’s free to use.
Who developed it: ChatGPT was created by San Francisco-based artificial intelligence company OpenAI.
Concerns: Last year, a Stanford University report said: “Models like GPT-3 can be used to create false or misleading essays, tweets, or news stories.”