Moore’s Law, coined in the 1960s states that computer chips will roughly double in power every couple of years. Sam Altman, CEO of OpenAI, the company behind ChatGPT, proposed a new version on Twitter last week that: “the amount of intelligence in the universe doubles every 18 months.” I’d like to suggest a third option; that the hype/hysteria (depending on your point of view) over AI doubles roughly 18 days.
Until that is the overexcitement just sort of fades away, either because the technology fails to live up to the hype – flying cars have been ‘coming next year’ for decades – or, much more likely, because we figure out ways in which it can enhance our jobs, not take them from us, like every good new tool.
And all the systems we are currently referring to as Artificial Intelligence really are nothing more than sophisticated tools because what they are not is intelligent, at least not in the way of being able to make truly independent decisions like you and me. ChatGPT is a Large Language Model (LLM), an algorithm trained on, basically, everything everyone has ever written on the internet pre-2021. Which is a lot given that there are more than 5 billion of us using it.
This training allows ChatGPT to understand naturalistic speech incredibly well, discerning our intent in a question in a way that just isn’t possible with a Google search query. That same training then allows it to answer us in a very natural, human way because LLM means ChatGPT is very, very good at determining word order based on what has been written before. It is basically the world’s very best predictive text generator and that makes it a usefully powerful tool. And yes, pretty gobsmacking in the right circumstances.
What that doesn’t make it genuinely creative though. Because ChatGPT learns and responds based on everything that has gone before it all its responses are, for now, averages or best guesses. In fact, this way of training AI has already come under fire for copyright infringement with Getty suing Stability AI, the company behind the Stable Diffusion AI image generator, for $1.8 trillion for using its photos without permission. Which raises the question of who owns the IP of anything created using AI. If you ask ChatGPT to write a script for a radio ad for you, does OpenAI own the copyright? Do the writers of every script it has scanned, analysed and averaged in order to produce one for you own it? Besides, “We want the most average campaign you can produce,” said no client ever.
Which is why, for now, ChatGPT and other MML-powered models, aren’t reasons to panic but should be a prompt to learn as much as possible about them. Because, like every other innovation that has come along and shaken up the creative industries – digital imaging, smartphones, photoshop, deep fakes – AI is an opportunity not a threat. Used well, ChatGPT can speed up the background research that goes into a pitch (but check its answers very carefully), it can compile a list of likely filming locations, it can summarise demographic and psychographic information in an instant. Think of it as the world’s most incredibly helpful intern.
Which leaves us more time for the irreplaceable, human bit of what we do, whether that’s building brilliant client relationships, having a lightbulb moment strategic insight, or coming up with genuinely original killer ideas. Something the current generation of AI just can’t do because it doesn’t create, it copies.
Is the level of true intelligence in the universe really going to double every 18 months, evolving AI from very clever predictive text and image generator to genuine creative genius? Well, we’re still waiting for our flying cars. Of course, I might be completely wrong, which was why I was very polite to ChatGPT when I asked it to write this article.



   digital
Share This