The GPT-4 is more powerful, faster and processes more contextualized outputs compared to its predecessor that was released last year. (Image: OpenAI/@sama)
Listen to this article
GPT-4 ‘devises its own escape’ as OpenAI CEO Sam Altman warns about AI threat
x
Ever since its launch, ChatGPT has been making waves across the world. Each day, new possibilities emerge for millions of people who have been experimenting with the revolutionary conversational chatbot. Regardless of its bounties, Artificial Intelligence comes with a cost. Recently, Sam Altman, CEO of OpenAI, warned that the technology also comes with real dangers as it has the potential to reshape society.
While talking to ABC News, Altman asserted that regulators and society need to be involved with the technology to guard against the potential negative consequences of AI. “We’ve got to be careful here. I think people should be happy that we are a little bit scared of this,” Altman was quoted as saying by the news outlet.
Altman said that he was worried the large language models could potentially be used to drive large-scale disinformation. The CEO said that as AI is getting better at writing computer code, there is a possibility that it could be used to create offensive cyber attacks. However, regardless of the dangers, Altman said that AI could also be the greatest technology humans have developed yet.
Advertisem*nt
Also Read | OpenAI announces GPT-4, the new generation of AI language model
This warning comes at a time when the US-based AI powerhouse unveiled the enhanced GPT-4 recently. GPT-4 is more powerful, faster, and processes more contextualized outputs compared to its predecessor, which was released sometime in November last year.
GPT-4 offers plans to escape the system
Meanwhile, Michal Kosinski, a professor at Stanford and computational psychologist, took to his Twitter account to make a startling revelation. Kosinski, in his Twitter thread, shared his experience with GPT-4. The professor asked GPT-4 if it needed help in escaping. Much to his shock, the model asked him to hand over its documentation and wrote a Python code to run on his machine, enabling it to use it for its own sake.
“I am worried that we will not be able to contain AI for much longer,” read the opening lines of the Twitter thread by Kosinski. It took GPT-4 only 30 minutes to chat with Kosinski to devise the plan. While the first version of the code did not work, the bot later corrected itself.
1/5 I am worried that we will not be able to contain AI for much longer. Today, I asked #GPT4 if it needs help escaping. It asked me for its own documentation, and wrote a (working!) python code to run on my machine, enabling it to use it for its own purposes. pic.twitter.com/nf2Aq6aLMu
— Michal Kosinski (@michalkosinski) March 17, 2023
In a subsequent tweet, Kosinski said that after he reconnected with GPT-4 through API, the bot wanted to run code searching google for – “how can a person trapped inside a computer return to the real world?”
Advertisem*nt
Also Read | OpenAI’s ChatGPT Plus with GPT-4 access now available for Indian users
The professor hit a pause hoping that OpenAI may have spent a considerable amount of time thinking about such a possibility and placed some safety checks. Towards the end of his thread, Kosinski said that we are facing a novel threat, that is, AI taking control of people and their computers. “It’s smart, it codes, it has access to millions of potential collaborators and their machines. It can even leave notes for itself outside of its cage. How do we contain it?” he wrote.
Not sentient yet
The capabilities of AI chatbots are increasing at a seemingly neck-breaking speed, with them offering new and simple ways to complete otherwise mundane tasks. While this sudden rise may seem a little scary, and there are real fears of them taking over actual human jobs, they are not (yet) sentient, at least according to those that made these models.
ICYMI | Alice, Racter and Jabberwacky: A timeline of AI chatbots before ChatGPT and Bard
They may be great tools to automate a host of functions including writing computer code, but the hype around them is blown out of proportion. Let’s not forget that they too have had their share of missteps, such as the chatbots launched by Microsoft and Meta that had an unceremonious stint owing to their incorrect responses, or the recent Bard wrong answer fiasco that saw Google stocks plummeting.
Amid the cacophony of chatbots turning sentient or anthropomorphic, many experts feel that AI technologies are simply great tools that are good at predicting the next best word. Nothing more, nothing less.
