Fake news and fake images will become some of the biggest
challenges with the proliferation of generative artificial intelligence (AI), which PitchBook estimates will reach $42.6 billion this year.
AI isn’t new. In December 1997, two computer
scientists, Sepp Hochreiter and Jürgen Schmidhuber, invented Long Short-Term Memory (LSTM) networks, which improves memory capacity in neural networks and allowed for faster and more precise
pattern recognition in training data, according to Bryan House, chief commercial officer
at Neural Magic.
But through the years, companies like OpenAI, Google, Microsoft, and Meta have developed technology to advance neural networks. And it wasn’t until 2012 when
several major breakthroughs happened for the performance and accuracy. For example, former head of Google China’s Kai Fu Lee, argue that the last great innovation for deep learning happened in 2012.
Perhaps until now. "60 Minutes" saw the potential of the
experimental text to video AI during an interview at one of Google’s campuses. Eli Collins, vice president of research at Google, demonstrated how the prompt “Golden retriever with
wings” created pictures out of words.
There are safety filters for the technology. For example, it does not create images of people. Google CEO Sundar Pichai told "60 Minutes" that
someone could create a video of him saying something. He may not have said it, but it looks like it.
“There are deeper concerns people worry about,” Pichai said. “At some
point does humanity lose control of the technology it's developing?”
One of the most interesting things about Bard revealed by the interview with "60 Minutes," which aired Sunday night,
Bard does not look for answers on the internet like Google search. Bard's replies come from a "self-contained program that was mostly self-taught," with microchips "100-times faster than the human
brain."
James Manyika, SVP of technology and society at Google, told "60 Minutes" that it took Bard more than several months
to read most everything on the internet and create a model of what language looks like. Bard, however, can not think or make judgements.
Despite concerns, the generative AI market is
expected to grow at a 32.0% compounded annual growth rate (CAGR), reaching $98.1 billion by 2026, even without accounting for the potential of generative AI to expand the total addressable market of
AI software to consumers and new user personas in the enterprise, estimates PitchBook.
New business opportunities will rely on user experiences. The advertising and technology industries have
already seen commercially successful products based on generative transformer and score-based diffusion models. And it’s estimated by PitchBook that solid precedent exists for future research
and development to yield results within just 18 to 24 months.
It will become crucial to ensure that foundation models have guardrails in place to guarantee trustworthy outcomes. For startups
looking to take on industry incumbents, the keys to success will be user experience, customization, and access to proprietary data, according to PitchBook.