ChatGPT CEO admits he is ‘scared’ the bot could be used for ‘large-scale disinformation and cyberattacks’
- OpenAI CEO admitted he is scared about what humans will do with ChatGPT
- Sam Altman spoke with ABC NEWS about the bot’s abilities and downfalls
- READ MORE: Everything you need to know about ChatGPT 2.0
OpenAI CEO Sam Altman admitted his is scared about ChatGPT’s abilities, but mainly with how humans will use it
OpenAI’s CEO revealed he is ‘scared’ that ChatGPT could be ‘used for large-scale disinformation and offensive cyberattacks.
Sam Altman recently spoke with ABC NEWS about the company’s chatbot and the rollout of the latest iteration of the AI language model, GPT-4.
While the chatbot has sparked fears of AI world domination, Altman sees humans as the greatest threat to the technology.
‘There will be other people who don’t put some of the safety limits that we put on,’ he told ABC News.
‘Society, I think, has a limited amount of time to figure out how to react to that, how to regulate that, how to handle it.’
OpenAI launched GPT-4 last week, touting it as more powerful than its predecessor – so much that it could be ‘harmful.’
The bot accepts inputs in the form of images as well as text, but still outputs its answers in text, meaning it can offer detailed descriptions of images.
The company warned that the model is still prone to ‘hallucinating’ inaccurate facts – and can be persuaded to output false or harmful content.
‘The thing that I try to caution people the most is what we call the ‘hallucinations problem,” Altman said.
‘The model will confidently state things as if they were facts that are entirely made up.’
While the chatbot has sparked fears of AI world domination, Altman sees humans as the greatest threat to the technology
During the interview, Altman admitted that GPT-4 is ‘not perfect,’ but can write computer code in most programming languages and can pass the Uniform Bar Exam in the 90th percentile
However, Altman sees GPT-4 going rouge in the hands of humans who use its powerful for evil.
READ MORE: GCHQ spy chiefs warn Microsoft’s ChatGPT and other AI chatbots pose a security risk
In a blog post, the National Cyber Security Centre says companies operating the technology – like Microsoft and Google – are able to read questions typed into the chatbots.
‘I’m particularly worried that these models could be used for large-scale disinformation,’ Altman told ABC NEWS.
‘Now that they’re getting better at writing computer code, [they] could be used for offensive cyberattacks.’
He also touched on another fear spreading across the world – AI taking over human jobs.
Altman is not blind to the fact that technology will replace some careers and believes it could happen faster than we think.
‘I think over a couple of generations, humanity has proven that it can adapt wonderfully to major technological shifts,’ Altman said.
‘But if this happens in a single-digit number of years, some of these shifts … That is the part I worry about the most.’
GPT-4 is available now to subscribers via ChatGPT: free users will still experience the older GPT-3.5.
Speaking this month at SXSW, OpenAI cofounder Greg Brockman said that fears of AI tools taking people’s jobs were overblown – and that AI would free up humans to focus on important work.
‘The most important thing is going to be these higher level skills — judgement, and knowing when to dig into the details,’ he said.
‘I think that the real story here, in my mind, is amplification of what humans can do.’
The new features of GPT-4
The big new feature here is that GPT-4 can respond to visual prompts – so for instance, it can reply to, “What’s funny about this picture?”
Longer prompts and responses
GPT-4 can now accept prompts up to 25,000 words in length, and output text of a similar length.
It can learn your writing style
GPT-4 is capable of learning a particular writing style, which makes it an ideal partner on creative projects.
Better exam results
GPT-4 now places in the top 10% for exams such as the Uniform Bar Exam.
It’s better at getting its facts right
GPT-4 is 40% less likely to come up with factual errors – although it still does so – and is 82% less likely to come up with banned content.
It can interact with web links
You can now ask GPT-4 to interact with content from web pages.
Source: Read Full Article
Satellite images show landmarks from above – how many can YOU guess?
Here's what the home of the future might look like, according to AI
Phones that emit the most – and least – radiation
Weirdest items found inside old furniture – including gold teeth and dead lizard
Model snaps selfie on beach and accidentally catches woman ‘taking photo of her’
Woman bounced back from death to spread news of the ‘afterlife without time’
Can you tell what it is? Rare 'fungus' on display at Kew Gardens
‘I can see into the future – losing control of civilisation should terrify us’