It's always exciting to see how AI is being used to enhance the next generation of products. This is a recap of my CES 2018 experience through the lens of a developer with a focus on everything AI.
I've attended CES for the last 4 years. This time it was a little different, this time I was attending with a deeper knowledge of AI. This time, products that used AI had my full attention and I focused on those booths.
What I like about attending CES is that it's a nice break from typical developer conferences. At CES, the entire packaged product is in full view. The technical side of me enjoys the perfect marriage of software and hardware. But the entrepreneur side of me enjoys the way the companies package the tech and market it to the masses. This is especially interesting when it comes to AI.
It was interesting to see different ways to package complex technology. Companies needed to simplify the tech to something the consumer could understand. I found the most difficult part of product creation was the messaging. Communicating the value to the end user is not always straightforward. This was the Achilles heel of my last startup but that's a long story for another post.
With an eye on more than the tech, I came to an interesting conclusion about productizing AI:
"It seems that pure AI products fell short of expectation. Products which focused on delivering AI failed to show tangible value to consumers. But products that used AI to enhance non-AI features were able to take a more realistic approach to AI"
Let's go through some of the highlights and I can give you more details on why I came to this conclusion.
6 groups of AI at CES 2018
This wasn't the way CES grouped the tech. There was so much to see at CES, any formal grouping would be a futile effort. There's so many ways to group products. There's no right or wrong way.
If I was to only include products related to AI, I could group products in the following categories:
- Auto Industry
- SMART everything
Some of the above categories did better with AI than others. There were really cool products that I wanted "now". Other products did so bad they should have just stayed home.
Not all products called out AI in their messaging. Others like virtual assistant and robots didn't have to call it out since it was a given. Finally, there were products that highlighted their use of AI as way to be hip or to ride the hype train.
The auto industry rocked AI
This is an obvious category that has pioneered the use of AI. It's hard not to lookup AI and not come across autonomous vehicles and all the AI that goes into them. If you're interested in AI + autonomous vehicles checkout drive.ai.
Autonomous vehicles was a key topic at CES 2018. CES 2018 also highlighted a good amount of new features which didn't relate to autonomous vehicles. These features related to enhanced experience for the the human driver. For example, a car could disable the accelerator if it detected a potential collision. Or, it would lock the door if it detected a pedestrian on a bike approaching from the driver's side. This is where AI really shined beyond driverless cars.
"AI kicks butt when it doesn't replace the human but enhances the human's interaction"
There was one really cool exhibit that was a pure play on AI and did replace the human. Yamaha took a top of the line racing motorcycle and added a robotic driver to make it autonomous. Yamaha claimed that their robotic driver can beat the best human racer. I believe it! but man does it also look so cool.
No doubt AI is killing it in the auto industry. The value AI can create is very clear to consumers and they're anxious to get their hands on it. I didn't spend too much time on this group since I get enough of this from online media.
AR/VR, the hidden AI
This category was a little more subtle on how AI was being used. The products in this area didn't directly call out the AI components of the product. It took a little bit of chatting with the right person behind the table to get to the AI parts. I spent most of my time looking at Augmented Reality (AR) and skipped much of the Virtual Reality (VR).
I didn't see anything exciting in VR expect for one product and it had nothing to do with AI. It was an 8K resolution VR headset from PIMAX boasting a 200 degree FOV! This was one of the products I really wanted to test but ran out of time.
As for AR, there were few companies doing cool object and surface detection. My impression of the demos was so-so. The clarity of the AR images was still off and too transparent. The stability of the AR images was good but most demos required markers which to me is cheating a little.
I was hoping to see more innovation in the computer vision + AI software driving the headsets. The focus was more on the quality of AR image than the utility of the product.
Reflection is a typical way for AR headsets to overlay an image in your filed of view. Reflection on a transparent surface is a good way not to obstruct the view in front of the user. Think of a typical heads-up display projected on a car windshield. Using this method to create AR, gives a very weak/transparent image and limited depth of view.
I did like one reflection-based AR headset that allowed the user to use their smart phone to create the AR experience. Mira is aiming to makes AR headsets more affordable. This is comparable to the approach Samsung GearVR took where a Samsung phone fits into the VR headset. The AR experience was not the best but if priced right, it could offer consumers an entry level headset into AR.
What did get better from former years was the aesthetics, size and weight of some of the devices. This shows that the industry took notes from the failed Google Glass experiment. The industry knows that there is only so far people will go when it comes to looking weird for the sake of better tech. Solos displayed a good looking set of glasses that provides runners and bikers heads-up displays with style. The glasses from Solos looked like a next generation Google Glass headset.
The most impressive piece of hardware in AR category was from LetinAR. The AR glasses from LetinAR don't use reflection. This new device actually had a special lens that projects the light directly into the eye. You might think at first, I will never let something shoot light into my eye. But when you think about it you are letting your laptop or phone do that same thing right now but at distance. These lenses are under an inch from your eye so the emitters can be small and not obstruct the view behind the lens. I tried it on and it was impressive.
The FOV was close to 80 degrees vs 50 that you experience with other high-end devices. I tried the HoloLens in another conference. On HoloLense, I was very impressed with image stability but not the FOV.
The quality of the image was also very good. The image was very solid with little bleed through of the background. This size and comfort level was also very good despite all the onboard hardware. The phone can run disconnected or pair with another device like a phone.
"I see non-reflective based AR the longterm winner. Devices like LetinAR and magicLeap are going take over soon. Devices that use reflection will loose ground and need to adapt fast to stay in the game."
I know I talked more about AR rather than AI in this section. This shows how far behind this industry is with productizing AI. This is the case even when AI is very strong in things like image recognition which is core to AR.
OMG! It was crazy how many players are in this space. It seems like every company wants to connect their product to the internet and have you talk to it over an app. In my previous post on learning AI, I mentioned IOT and how its gonna drive the AI industry and the jobs in that space. CES 2018 sent a clear message that IOT is huge and just gonna get bigger.
The biggest industry taking advantage of this trend was the smart home industry. It seams like there are sensors built into every piece of hardware you can install in a home. It has gone beyond the thermostat and garage door. You can track individual power outlets, fire detectors even water leaks in bathrooms. Of course, the big thing now is "connected appliances". Appliances that talk to the web and tell you things. Your fridge can tell you when you are low on milk. Your water heater can tell you when it needs maintenance.
I didn't come across much onboard AI in the IOT space. AI played a role in the apps mining that data produced by the IOT devices. There was one startup at Techstarts Startup Corner that blended AI and IOT. They called it AOT "Autonomy of Things"
"AI + IOT? hum... I wonder if I'm witnessing the seeds for what is to become a future depicted in the movie "Terminator" :)"
This part is where I rant a little bit about poor demonstration of AI and the other hyped use of AI. VA Chatbots and Smart Devices trying to make decisions for you are big fails.
Unfortunately, this category had the largest AI presence at CES 2018.
There was a meme that was floating around on how GE's demo of a smart home assistant failed mid demo. This meme set the tone for the tech in this space and my experience matched that tone.
I was walking through the smart home section and came across a group laughing in front of a large TV screen . I peeked over at what was on the screen and noticed a chat interface. The promise was that you can use a chatbot to help check and control your home remotely. I didn't see the entire chain of messages but was able to capture the bottom part. The ridiculous responses were amusing though.
This seems to be a trend in the chatbot space. Many large companies have decided to scale back their ambitions in Chatbots. The tech is not good enough yet and will not be for a long time. There is a good reason for this.
The problem with VA Assistants is that there are two sides of the problem that need solving. The NLP (Natural Language Processing)/Voice Recognition is one side. The expert system that makes the decisions on how to solve your question is the other side.
There has been amazing progress on the first side of recognizing the request using NLP. The logic on the backend is still far from being practical.
There are some newer AI chatbots where it's almost impossible to detect if it's a human or AI agent. This is the good of chatbot tech.
The bad comes from the moment the AI agent needs to figure how to solve your problem. The agent would have to reach higher levels of "General Intelligence" for this to reach the level of a human. This is where the tech falls on its face.
"Artificial General Intelligence is still in its infancy."
The CES conference app also had a chatbot. I tired it several times when I could not find what I looking for in the app. I got responses that were links to irrelevant web pages. I stopped using it after day one.
"I do believe that Ai can play a significant role in the VA space but not as a pure AI solution. If you keep the human in the loop and just use AI to enhance the interactions, you will win with AI. "
Another company hard to ignore at CES, was Google pushing their "Google Assistant". This is Google's answer to Amazon's Alexa.
Google had reps running around in white jump suits with their logo on the back. Google also had reps at booths of smart devices that integrated with their platform. They also had huge displays with giveaways and challenges with prizes.
I don't know why but it annoyed me. I felt that the tech can't sell itself so they are pumping tons of marketing dollars to get us to bite.
I have asked many friends and family, that have used Alexa, what they use it for. The answer is always the same; music and checking the weather.
Not very impressive value add from sophisticated and expensive AI.
I know that companies are still gonna still push hard to adopt chatbot tech. It only makes sense to reduce customer support staff even if it doesn't improve the overall customer experience. They did it before with IVR's (Interactive Voice Response) and they will do it again with chatbots.
Robots, the mechanical humans.. NOT!
This area is the most pure implementation of AI. Meaning, AI is closest to the core value prop of the product. This group of products were still in the early stages of development. Robotics have a ways to go before they deliver real value to the general public.
Industrial automation is where robotics are making this biggest impacts. Robots used for security monitoring in large facilities has also shown good results.
Flakey demos were common in this group. Robots didn't respond correctly to human voice commands. The response were slow and difficult understand. Some booths claimed amazing features but didn't have the robots functioning.
In general, I didn't see that home or even business based robots worth the investment yet except in very limited use cases.
The coolest of all the robot demos was a ping-pong playing robot that played against a human player. The robot was capable of playing at very high skill levels. It was also capable of learning the skill level of the human player and adjust to help train the player. The robot would let the human win every now and then if it meant improving the player's skill. It was a really cool demo but not sure there was a huge market for that kinda robot.
Show me me the data
The picture above was from a live demo where the cameras where detecting faces real-time and trying to match them with profile photos of attendees. It's almost impossible for a human to track all faces crossing the frame of a video screen. AI on the other hand handles this with ease.
It was a little spooky to watch my face get tagged and then the system run through similar faces in the database. Was clearly "big brother" tech, but I can see good real world needs for solutions like this.
I think the biggest win in industry leveraging AI is in the data analytics space. Using AI to help humans mine and track large amount of data is AI's biggest value add.
"AI runs on data. The larger the data problem, the stronger the AI solution."
And the winners in AI are...
CES was fun but a little overwhelming, I'm sure I missed a bunch of AI based products. There was way too much for a single person to see. I did however see more than what I listed in this post. I listed only the AI products that grabbed most of my attention and was worthy of mention.
It was clear to me that some categories were implementing AI much better than others. Products that blended human behavior with AI assisted behavior shined the most. Examples of this were in the Auto, Data and AR categories. The Robotics and Smart Device categories focused on replacing human behavior. These categories failed to show enough value for me to bite. The growth in the IOT category cemented the need for AI. The future will bring a huge amount of data from IOT and we will need smart AI agents to make sense of it all.
I hope this was a fun read for you. If you have any comments, let me know below.
If you're a developer interested in getting started with AI or just curious about AI, checkout the mind-maps I put together. There's lots out there on AI and if you're like me and like visuals, you will love these maps. (click here)