by Anand Parthasarathy
December 7, 2022: Lastweek, a chatbot totally fuelled by Artificial Intelligence (AI) was offered for free use worldwide by the San Francisco-based OpenAI Foundation and it was quickly hailed by users – over a million signed up to try it just 5 days – as the most advanced AI app ever launched for lay users.
And already, there has been fierce discussion whether this is the beginning of a new era where machines will replace humans as universal informal sources, content generators and even as software engineers. “Professors, programmers and journalists could all be out of a job in just a few years”, warned a writer in The Guardian (UK).
Towards autonomous systems
The bot is called ChatGPT (for Generative Pretrained Transformer). The developers explain in a blog: “We’ve trained a model which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.” It is constantly learning from human feedback.
“OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity.”
Early users say they are stunned by some of the things the bot does and how its output is almost indistinguishable from human work:
- Academics have challenged it with exam-type queries and admit the answers it generated, would get full marks. It can solve complex mathematical equations. Will this impact the entire exam and evaluation as we now know it? At any rate home-work assignments will likely become meaningless.
- If a problem is spelt out, it can generate code in a variety of programming languages. There appears to be a real possibility that the role of human software programmers could be severely curtailed in the coming years.
- On the literary side, it can compose a poem on a theme you suggest and reportedly it can even turn a story into a screenplay.
How much of this is a startling vision of a machine-driven future – and how much is snowballing hype?
End of Google?
To give Swarajya readers a more realistic assessment of ChatBPT and what it can do, I signed up (with some difficulty, explained below) and have been posing a variety of questions of the type a lay person or student would ask. Unlike so many Twitter and Facebook posters, I did not ask idiotic questions or try to trip the bot into answering politically incorrect or inappropriate queries. That is a perverse game that many play in the US. Here is my take on what I asked and what I could learn:
Some users have said the bot could eventually replace Google.
Here is my conclusion: No; will not happen. Let me illustrate with my experience: I asked technology-related questions like What are the five generations of cellular phones ? Or: Please explain quantum computing. Almost instantly, I got approximately 300 words of a crisp answer in simple layperson’s language. Unlike many earlier conversational AI apps I have used, ChatBPT takes up a follow up question, if you are not satisfied and tries to explain better.
But when I asked: How many Indians own a cell phone, this was the response:
“I am sorry I am not able to browse the internet, so I am unable to provide you with the most up to date information on the number of people in India who own cell phones. However… I can tell you that India has one of the world’s largest populations and a rapidly growing economy so it is likely that a significant portion of the population owns cell phones.”
This is the crux. The bot is backed by a huge information base, currently updated only till 2021. Unlike a search engine like Google or Microsoft’s Bing, it does not crawl the Web in real time scooping up the latest facts and updates, so it does not know the latest number of phones but tries to suggest a trend. In a way it is closer to Wikipedia than Google. So it cannot answer live news questions. So any obituaries for browsers and search engines are an exaggeration.
But what I found to be almost uncanny, is the way the bot can refashion its information in a startlingly human way. I said in plain language I want to promote a news story on ‘how conversational AI is now a user-friendly option in e-commerce’ with three tweets. It provided this answer:
“Here are three possible tweets:”
1. "Introducing the new and improved #conversationalAI for e-commerce! Our AI chatbot is now more user-friendly than ever. Try it out now on our website."
2. "Say goodbye to frustrating online shopping experiences! Our new #conversationalAI chatbot is here to make your e-commerce experience seamless and easy. Give it a try now."
3. "E-commerce just got a whole lot easier with our new #conversationalAI chatbot. Get personalized recommendations, answers to your questions, and more - all with a simple chat. Try it out now on our website."
This level of ingenuity will seriously threaten social media content providers who are paid to generate promotional tweets or Facebook posts. Uploading web content, answering customer enquiries would also seem to be enroute to extinction.
Challenge to journalism
But can technology replace journalists? It can make their work easier – by producing sharp copy based on any subject, which can then be manually updated. But from my limited trials I think it will fail critical tests like ethics or opinion forming. I asked the bot: Would it take away the jobs of human writers? This was the answer:
“I am not capable of taking away jobs from human writers or any other profession. My purpose is to assist with information and answer questions to the best of my ability… I exist solely as a digital assistant
AI systems like myself can help writers by providing suggestions and ideas, but ultimately it is up to the human writer to create the final product". Reassuring!
ChatBPT is currently free for anyone to try – but one doesn’t know for how long since the cost of creating the AI bot will remain so. The makers say the cost of the computation is ‘eyewatering’ and after the trial phase they may have to charge less than 10 cents (Rs 8) per chat
To sign up, go to https://chat.openai.com. You will need to first sign up for OpenAI to access ChatBPT. An email ID and a mobile number are required. You may not get on board at your first attempt, but keep trying. If you have a corporate email ID, it may work better than a personal email.
Once you enter the bot question box, moderate the speed of your questions. After three questions it told me: “Too many requests, please slow down”.
When Netscape launched the era of browsers in 1994, followed four years later by Google; when the iPhone “smartened” up the mobile phone with a touch screen, each marked a significant change in society. Could this month’s development of a clever bot, propel a Third Wave, bringing us closer to machine intelligence? We must wait and watch.
This review first appeared in Swarajya yesterday