Last Month in AI:
Slowdown?

Let’s take a quick look at what happened in November to check for a slowdown. Is it real or just the latest media narrative confected to get people clicking? Judge for yourself:

• Coca-Cola dropped an AI-generated Christmas ad that scored five stars in effectiveness testing, even if old school (m)ad men hated it. Just five stars? If we’re measuring AI’s creative prowess, shouldn’t we be inventing new stars by now?

• Fashion brand Mango joined other brands in ‘supplementing’ human fashion models with AI-generated models to showcase their latest lines. Even if AI models have a lower carbon and financial footprint than their human predecessors, they’re hardly breaking news.

• In November, “Now and Then,” an AI-restored track from the Beatles, snagged a Grammy nomination, while the first AI-generated artwork – a portrait of Alan Turing, called ‘AI God’ from robot artist Ai-Da – sold for a seven-figure sum. One Grammy and a single million-dollar piece? So much for an AI revolution.

• The latest State of Generative AI in the Enterprise report revealed AI investment has increased by a mere 800% over 2023 levels. Over at Google, only a quarter of their code is now AI-generated. The very definition of a slowdown.

• A new MIT study found that AI-assisted researchers only make 44% more discoveries, leading to a mere 39% increase in patent filings and a paltry 17% rise in downstream product innovation.

• In China, a Bain study found that only 52% of e-commerce merchants are using AI tools, with a mere 75% seeing significant sales impact. Disappointing.

• Looking at job impacts, HBR reported that writing jobs have decreased by just 30% post-ChatGPT – at least on a major freelancer platform. Software development gigs dropped by only 21%, and graphic design by 17%. The robot takeover seems to be running behind schedule.

• The Journal of the American Medical Association reported that ChatGPT outperformed human doctors at diagnosis again, but doctors working with ChatGPT barely outperformed their AI-free colleagues. Message to humans: forget that fuzzy AI-human collaboration – just let AI do its thing.

• A study published in Nature pitched neuroscientists against LLMs, asking both to predict the outcomes of novel experiments. Team LLM won, getting 81% of predictions right, compared to 66% for human neuroscientists. Just a 15% improvement over human experts? Not impressed.

• A final study, this time out of Stanford, found that chatting with an AI for two hours gives the AI all it needs to clone you – your words, your personality, your values, your biases with 85% fidelity. British political strategist Dominic Cummings believes we’ll be polling and surveying AI clones rather than busy distracted humans in the future. But come on – two hours to clone someone’s personality, preferences, and biases? It’s 2024. Should be immediate.

All this non-news shows AI has indeed hit a wall, its bubble has burst, and we’re about to enter a new AI Winter. It’s all so disappointing. Right?

Given that nothing of any note whatsoever happened in Generative AI in November 2024, the AI slowdown must be real. Brands and their agencies can heave a sigh of relief. It was all a bad fever dream. AI is neither a promise nor a threat. You can safely ignore it. Move along, folks, nothing happening here – just the casual reorganization of human civilization, one “disappointing” breakthrough at a time.