Listen to this story
|
Slowly but surely, the buzz around ChatGPT-like chatbots is nullifying. At an event last week, Sam Altman spoke about how GPT-5 is not in the works for some time. Sundar Pichai, in a recent interview, said Google is holding back from releasing advanced versions of Bard – a ChatGPT rival – as a way to get “user feedback” and test out more safety features before releasing them.
Could the tech giants seemingly slowing down on releasing advanced capability chatbots, be an indication to AI development flatlining or plateauing after hitting its peak, particularly in the case of LLM-powered chatbots?
Safety first
The reality is, with problems such as hallucinations and non-factual data, chatbots are far from perfect. And with rising safety issues and security breaches that have led to instances of bans etc, the rise of chatbots seems to have hit a roadblock. Organisations are now considering safety before proceeding, which implies a pause in new development.
Sundar Pichai in a recent CBS interview talks about how Bard has included a “Google it” button to address hallucinations. Google has also built safety filters into Bard in order to filter out “hate speech” and “bias”. In addition to these features, Pichai also talks about AI system alignment with human values. He believes that developing such a form of alignment, should include social scientists, ethicists, philosophers and not just engineers. Society will play a huge role in figuring things out, “it’s not for a company to decide”.
Google released its AI chatbot Bard a few months ago with the company saying that “external feedback with internal testing” will help them improve Bard’s responses in order to achieve better quality and safety. Google’s rushed affair in the AI chatbot space was evident from Bard’s incorrect answer at the Paris AI event, a fiasco that cost Google more than $100 billion in market value with stocks falling nearly 8%.
OpenAI’s ChatGPT, on the other hand, that started the chatbot rage, since its launch in November, has been following a different approach. Releasing the latest version of GPT-4 that powers ChatGPT Plus, OpenAI has been on a continuous growth, however, laden with problems, nonetheless.
Progress later
An open letter was signed by over 25000 people including Elon Musk, Steve Wozniak and Gary Marcus to pause training of models beyond GPT-4 for six months, owing to “profound risks to society and humanity”. The move, though debatable, has triggered a reaction among corporates to address security concerns on priority. Referring to the multiple safety issues, OpenAI is of late on a spree to address AI safety. There have been a total of three statements lately, either in the form of company blogs/statements or Sam Altman himself talking about the future path for the company.
Altman even emphasised on how the company will move forward with caution and increased rigour when it comes to safety, also implying that there would be no newer models for the time being.
Less Noise, More Progress
Far from the race is Meta, which has always been a laidback player in the LLM race, besides Apple. Slow to the chase, Meta announced LLaMA, a foundation language model that ranges from 7B to 65B parameters, only towards the end of February when others had already set sailing the chatbot way. Though Meta had forayed into LLMs with past ventures such as BlenderBot and Galactica, poor performance with hallucinations and racist responses led to its failure. Maybe, Meta learnt from its past failures, and is currently focusing on what’s working for them and its users.
Irrespective of the slowdown in new chatbot releases, Microsoft has been the biggest winner with ChatGPT powering their search engine Bing, and integrating the same on their cloud computing partner and enterprises. The recent news on how Samsung will replace Google with Microsoft’s Bing Chat as a default search on their devices led to Alphabet’s stocks crashing amounting to a loss of $57 billion, reflects on how a chatbot can determine supremacy in the market.
Though chatbots and LLMs are being used across industries, going by the statements and interviews given by the leaders of OpenAI and Google, it is evident that there would be a certain slowdown in the chatbot race.
Apart from improving safety and security features, the companies might not release anything new in the immediate future. If not an end, the chatbot buzz seems to have subsided – Meta is handling it smartly, while OpenAI and the rest are going with the flow. But, the question is for how long?