A few weeks ago, Cadence hosted an interesting panel discussion that talked about how AI is going to impact various industries. Panelists included Prosit Mukherjee from Qualcomm, Sanjay Gupta from NXP, and Rajbir Singh from MediaTek, and was moderated by Jayashankar Narayanankutty from Cadence.
There were many topics that were discussed, including computing on the edge, contextual awareness, and security. Since there has been a lot of news about 5G lately, in this blog I have focused on what the panelists have to say about 5G in their fields.
Jayashankar (Cadence): 5G is finally around the corner. Tell us about the technology behind it and what is the impact that it will have.
Prosit Mukherjee (Qualcomm): The job of the 1G, 2G, 3G, and 4G networks was to provide reliable voice and data connectivity that was used to connect people. That was the job of the previous 30 years. But for the next 30 years or so what we should be looking at is not just connecting people, but connecting the world, which is what 5G is going to enable.
At this time, we are just scratching the surface of the potential. The rollout has been faster than what we had expected. In the US, Verizon has already started it. So, exciting times ahead.
Sanjay Gupta (NXP): Let me talk about autonomous driving and I’ll link it to the safety of human life.
Just to give you some statistics around that—roughly every 30 seconds, somebody dies on roads around the world, which roughly amounts to 1.3 – 1.4 million people annually who die due to automobile-related accidents. Another interesting fact is 90% plus of these accidents happen because of human mistakes. They are not because of electronics. So, the way we are addressing this application area is if we can enable artificial intelligence and super intelligence to replace all those areas where humans create those errors, can we actually save lives?
And in that spirit, we believe that autonomous driving is going to be the very big revolution to do that with what we call “sense, think, and act”.
“Sense” engine refers to the state-of-the-art radar, camera, lidars, and other sensors. So, sensors within the car and sensors outside the car to act like human ears and eyes, but fool-proof ones.
“Think” refers to the artificial intelligence part, where mimicking the human brain comes into the picture. We are at the tip of the iceberg here, but in time the electronics and semiconductors will make it possible to surpass the throughput, storage, bandwidth, and the latencies of all these electronics systems.
“Act” refers to the traditional electronics of actuators, whether it is the power train or braking or internal body gadgets.
From the semiconductor design perspective—data-generation capabilities, the data-transfer capabilities, the data-storage capabilities, are all are increasing complexity of chip design. Comparing the SoCs that we designed four years back versus today—the entire specifications have increased 100 times. So the entire capability of storage, of RAM, the cache—all these aspects that impact car-to-car communication—the data pipes have to be huge if we want to move into this field.
Eventually, the car is going to be almost like your Apple or Android applications to download. This is going to be possible with 5G.
Rajbir Singh (MediaTek): AI has been around for the last 50 to 60 years. We studied the algorithms in college. But let’s talk about a few trends that are driving the pervasiveness of AI today.
First, there is a lot of investment coming into AI. Every company is trying to get a piece of the AI pie.
Second, compute power/hardware. All these years, I think why AI held back is due to computing power and if there was computing power, it was among the privileged few. So the democratization of the computing power has helped in the realization of AI.
The third thing is the explosion of data. And billions of devices are producing so much data. The challenge is that all data is not labelled but that’s what is required by the AI algorithm.
Fourth, the availability of talent. There is so much investment happening in AI that it is attracting the best of talent in the industry. That is a challenge for all of us [in the semiconductor space], since it has led to a talent crunch in the market.
Fifth, applications. Every industry has so many applications whether it is automotive or healthcare or insurance. There are so many commercial, viable business models that need AI.
And lastly, algorithms. Like I said, the AI algorithms are the same as the ones we studied in college. Of course, there are now many modifications which companies like Google and Facebook have enabled for the normal user.