Key Takeaway
While some tech innovations are working to bring computer intelligence closer to that of the human brain, others are looking instead at how we can use computers to expand the brain’s existing capacity. Either way, humans and machines have never been closer.
Existing AI still faces a multitude of limitations when attempting to make sense out of unstructured data. Things like natural speech/voice recognition, language and video data comprehension are still in their infancy when it comes to matching the capabilities of the human brain. In order for machines to accurately evaluate these on a large scale, we need to continue to make advancements.
One thing our algorithms need more of? Data. Data and AI are tied closely together. AI needs massive amounts of data in order to learn and perfect its algorithms. Imagine all of the inputs a child receives on a daily basis as it begins to comprehend the world. Machines are no different – the data they’re exposed to allows them to learn from it, to discover patterns and trends and to recommend actions accordingly.
As we work toward a world where AI is more closely tied to the human mind, several startups are hard at work to improve data translation capabilities.
Graphcore helps tech ‘see’ video content more clearly
Even the artificial intelligence used by giants like YouTube and Facebook struggles to determine the content of videos uploaded to their platforms every day. As a result, associated ads and insights gleaned from them are often less than relevant.
U.K. artificial intelligence startup Graphcore is looking to change that with a chip that will power machine-learning algorithms that can process all the visual data uploaded to social media platforms, starting with servers processing data in the Cloud.
According to Forbes, the company raised $50 million from Sequoia Capital to produce a chip that can understand the context of and recognize images from video uploads. The team claims this hardware is 100 times more powerful than leading GPUs sold by competitor Nvidia and Google’s Tensorflow.
“Social media sites need to stay on top of the way people are communicating,” Graphcore founder Nigel Toon said. “It’s not just about text but images. What people are trying to think about is how to do video… If they don’t know what you’re doing on the social media site, how do they monetize the social media service?”
The company’s long-term goal is to allow all AI innovators to “ lower the cost of accelerating AI applications in cloud and enterprise datacenters to increase the performance of both training and inference by up to 100x compared to the fastest systems today.” It’s core product is an intelligence processor unit, or IPU.
Graphcore are not the only ones. Google, Apple, Tesla and Nvidia are all working to develop similar processors to suit their own needs. The resulting race might push new technology that will speed up the way we all process video data.
Social media sites need to stay on top of the way people are communicating
NLP takes a leap with translator technology Flitto
Once again, in an area where tech giants are all racing to develop their own technologies, a startup has taken the lead in helping us better process another area of unstructured data: natural language.
We’re all aware of the limitations inherent in Google Translate (how many of theseGoogle Translate fails are out there?). For those who rely on this to help them in business, these ‘fails’ can go beyond funny miscalculations to offensive, if not culturally insensitive, gaffs. It always requires human verification.
As an aggregate, we’re also still a ways away from accurately gathering data from many of the world’s multitude of languages in a culturally relevant way, especially when it comes to audio data. Most of the time the issue is simply not enough data to train the algorithms we’re using to translate.
Companies like ReCaptcha have been using crowdsourced human feedback to help contribute to this for years. Today, AI-powered translation startup Flitto intends to provide companies with the language data they need to train their machine translation programs. It began as a crowdsourced translating service, but now at a point where it has gathered sufficient data to train other programs, Flitto sells its language data, called “corpus,” to corporate clients.
According to Founder & CEO Simon Lee, what sets Flitto apart is that it offers “sets of human-translated sentences from its crowdsourcing service, which is used for things like slang, pop culture references or dialects that might stymie a machine translation service.” Over the last five years, Flitto has accumulated more than 100 million sets of translated language data.
We’re not yet at the point where machines can translate this data without the help of human verification. Once trained, however, they can help us process much higher volumes of this data than any human can. This is one area where humans and machines are becoming even more interdependent.
Using machines to improve the human mind: exploring neuroprosthesis
While we work to bring AI closer to mimicking the capabilities of the human mind, in other ways scientists and tech entrepreneurs are looking at using technology to improve the human mind.
Tech entrepreneur Bryan Johnson is looking to get inside our brains in an effort to “create a better human” with his new company Kernal. According to a recent piece in WIRED detailing his latest experiment on 25-year-old epilepsy patient Lauren Dickerson, he wants to do this by “building a ‘neuroprosthesis,’ a device that will allow us to learn faster, remember more, ‘coevolve’ with artificial intelligence, unlock the secrets of telepathy, and maybe even connect into group minds.”
This would take the form of a chip that connects to the brain through some yet-to-be-developed form of noninvasive interfaces that would allow us to connect the brain directly with computers in an effort to improve our memory capacity and help us to better understand, and thereby treat, neurological diseases. What’s more? Johnson intends to make this widely available for commercial use.
Interesting to note is that this is the same Bryan Johnson who started payments platform Braintree, which was acquired by PayPal for $800M. He’s not the only tech bigwig interested in this sort of research. Elon Musk’s Neuralink raised $27M to connect human brains with computers, ironically in an effort to “allow the human mind to keep up with fast-improving artificial intelligence,” according to TechCrunch.
What does all of this mean for the future of data? As we continue to improve the capacity of machines to match that of humans, and humans to match that of machines, one benefit we hope for is that technology will enable the processing (and creation) of ever more data, so we can be more efficient, make better decisions and ultimately increase our awareness of the world around us.
