In 2014, Elon Musk made it clear that he had some fairly strong opinions about the sorts of technology emerging in the 21st century. “I think we should be very careful about artificial intelligence,” he said. “If I had to guess at what our biggest existential threat is, it’s probably that. So we need to be very careful.”
A year later, he co-founded OpenAI, a for-profit organisation whose mission is to develop “friendly” artificial intelligence technologies.
Since then, the Tesla mogul has backed out of the company in order to pursue his plans for benevolent AI elsewhere – but OpenAI is still thriving. Under the control of its co-founder and CEO, Sam Altman, the company recently received a $1billion investment from Microsoft.
Satya Nadella, Microsoft’s chief executive, explained that the money wouldn’t all be handed over at once, but instead would probably be filtered in over the course of a number of years – perhaps as much as a decade. This makes sense for a number of reasons: firstly, because Microsoft is planning to see a return on their investment by commercialising OpenAI’s eventual products; and, secondly, because what OpenAI wants to achieve with the money isn’t actually possible yet.
“My goal in running OpenAI is to successfully create broadly beneficial AGI,” Altman explained.
AGI, otherwise known as artificial general intelligence, is – in layman’s terms – a machine that will be capable of doing whatever a human brain can do.
“The creation of AGI will be the most important technological development in human history, with the potential to shape the trajectory of humanity,” Altman said. “Our mission is to ensure that AGI technology benefits all of humanity, and we’re working with Microsoft to build the supercomputing foundation on which we’ll build AGI.”
But how likely is it that we’ll see AGI become a reality?
Advancements made in AI recently have been undeniably groundbreaking, and OpenAI has been at the forefront of it all. Earlier in 2019, the team’s technology managed to beat the world’s best players at DotA 2 – a strategy-based team-playing game that, just last year, AI was unable to win.
In such a short space of time, then, the technology has progressed enough to triumph over human minds in a game that is designed for human methods of problem-solving. That’s progress, for sure – but using a machine learning algorithm to win a game is worlds away from actually simulating the incredible complexities of human intelligence.
“We’re testing a hypothesis that has been there since the beginning of the field: that a neural network close to the size of the human brain can be trained to be an AGI,” said Greg Brockman, another OpenAI co-founder and the company’s chairman. “If the hypothesis is true, the upside for humanity will be remarkable.”
The key word here is if. Because it might not be possible. Maybe not within our lifetimes, maybe not ever.
But if it is – and if OpenAI are the ones to develop AGI, they have a huge responsibility on their shoulders. Making sure the technology is safe and ethically sound “is going to be one of the most important societal challenges we face,” Altman said – and he’s certainly not wrong.
Indeed, the so-called intelligence explosion is one of the greatest potential threats to human life as we know it, and one we get closer to the more we invest in AGI technology.
Even so, OpenAI is being realistic about their progression plans. They know that AGI is still something of a pipe dream, and so they’re focusing on much smaller projects with Microsoft’s investment in the meantime.
Just recently, the team built a system that enables AI technologies to better understand natural language patterns. As they explain:
“We’ve trained a large-scale unsupervised language model which generates coherent paragraphs of text, achieves state-of-the-art performance on many language modelling benchmarks, and performs rudimentary reading comprehension, machine translation, question answering, and summarization—all without task-specific training.”
This sort of system already has practical real-world applications; it can be implemented in digital assistants such as Siri and Alexa, utilised to automatically analyse written documents, or harnessed to create original content.
Again, this is an astonishing achievement – but it’s still just a drop in the ocean compared to the sort of technology that would be needed to simulate general intelligence.
Once upon a time, though, we would have said this feat was impossible. We would have said that beating DotA was impossible. Because we simply do not know what AI is capable of yet, and the only way we’ll find out is if companies like OpenAI and Microsoft keep pushing for change and progress.
So, is it possible? Maybe. Should we pursue it if it is? Again, maybe. We’re still only at the dawn of AI’s potential – and waiting for the results may take some time.