Written by: Bennett Falck, Head of Product Development at Green Cabbage
When OpenAI released Chat-GPT to the public in late November of 2022, it immediately broke records for user adoption.1 In just these past eight months, Chat-GPT has transformed many people’s daily lives to a point where they are starting to forget what life was like without it. From marketing content to code review to general use as a simple search engine, Chat-GPT kicked off an AI revolution that was suddenly accessible to people with limited technical skills.
Technology is always an arms race. People will continue to innovate and improve products in search of profit and market share; that’s the beautiful thing about competition. A few Chat-GPT competitors have come out since its launch, including Google’s Bard. Launching worldwide in May of 2023, Bard experienced some issues upon full user expansion and is still considered well behind OpenAI’s offering both in terms of use and response quality.2 Given Microsoft’s heavy investment into OpenAI and Amazon’s existing strengths in AWS, this has led many to speculate that Google may be panicking internally.3 After all, if Chat-GPT can be a far better search engine than Google and AWS continues to outpace Google Cloud (GCP), where does Google fit? I believe the answer is already in Google’s hands, and it may revolutionize the world even more than the widespread popularization of AI.
Google researchers recently announced their latest sci-fi seeming creation: a 70-qubit quantum supercomputer. According to estimates, tasks that would take the world’s current strongest supercomputer 47 years to complete can be done by this machine instantly.4
So, what is quantum computing? Originally making headlines in 2019 when Google researchers announced that their latest quantum supercomputer had surpassed the best supercomputers in terms of capabilities and speed.5 In the four years since, they’ve increased that advantage by a factor of more than a billion by introducing more qubits. Qubits are like the binary 0’s and 1’s familiar to those with some computer science knowledge, except that they can exist in both states simultaneously, like a light switch that can be both on and off at the same time depending on needs. Quantum computers use this unique power to perform processes in parallel with other processes, greatly increasing the speed to completion of the request.
Now, what does this mean? While we are still at the early stages of quantum computing development given the delicate nature of qubits and extreme capital cost of creating such machines, these developments promise to unleash capabilities far greater than what we see today. For example, the latest release of Chat-GPT, GPT-4, was estimated to take 34 days to train on over 1 trillion parameters.6 Using Google’s new quantum supercomputer, this would have taken just milliseconds, and could incorporate many more times the original trillion parameters for continued response quality improvements. Simulations that are currently regarded as too complex, such as global solutions to climate change or new world-altering innovations, will become commonplace. While these quantum computers are unique in that they are generally engineered for a specific task or set of tasks, meaning they likely will never be consumer machines, these computers have the potential to revolutionize AI and machine learning in ways that we are not yet fully capable of conceptualizing.
Though Bard was rushed to the market and has largely failed at capturing a user base in the way that Chat-GPT has, Google is still at the technological forefront, just not necessarily where you might expect.
References
Opmerkingen