The AI folks at AI In Brief X, the folks behind the former Twitter, have been teasing their new AI product that’s been given the nickname Grok. They reckon it takes influence from Hitchhiker’s Guide to the Galaxy and it’s super-duper smart. Elon Musk, who now owns the AI company, boasted about it on Twitter saying it’s the best AI that currently exists. We don’t have access to it yet, so who knows how it is getting on.
Turns out that big language models can be faster than humans at composing phishing emails, but the writing is not as convincing as that through some good ol’ human trickery. IBM’s security bod, Stephanie Carruthers, ran an experiment and discovered her team could produce better scammy messages than the AI. They said AI is not quite as good at social engineering because it lacks the emotional understanding needed to manipulate victims. So, despite it being faster, it isn’t as good yet.
In a sort of similar vein, the big AI companies like OpenAI and Google DeepMind have signed some agreement allowing governments to test their AI models before they are pushed out to the public. This is due to all the worries about AI being used for dodgy stuff like spreading misinformation and disrupting jobs. The legal stuff isn’t clear so hard to know whether progress will be made. Will see.
And there’s a bit of a debate about AI training datasets taking info from news articles. It’s worried that chatbots are basically regurgitating copyrighted news content without permission or compensation. So, there ya go.