Artificial Intelligence and Machine Learning explained

DrHow

Going “Plaid” ASAP (Tesla S) RDT refugee
Advanced User
Lifetime Premium Member
Joined
May 18, 2018
Messages
2,543
Reaction score
4,031
Only device out there I can think of that does edge inferencing is Tesla autopilot. Definitely in a vastly different price range.
It is/will be used in Cyber Security.
Post automatically merged:

Speaking of AI... Will my Radenso RD be able to learn on its own on the road? OR is the AI "in house" where Radenso employees train Rai, and then those trained results go to my detector as an update? Or maybe it can go both ways?? Maybe Rai see's a new signal it has never encountered before and those results could be uploaded to Radenso to analyze?

Though I have a degree in I.T. and have been working in various I.T. fields going on 20 years, I really don't know much about the fundamentals of AI. I'm guessing data can be shared both ways between the detector and Radenso. I know the hardware on these new detectors will be very "beefy" compared to "legacy" detectors, but I wasn't sure if this new hardware had the capability to learn and adapt on the road - I'm thinking it does.

Are there any challenges you face dealing with airflow/cooling with the new hardware? Do you think there will need to be some kind of active cooling as opposed to today's passive vented cases?

Hope that makes sense..

Thanks,

David
The term “Field programable gateway array” FPGA has the word field, as ”in the field”. :) Serious, the ultimate use of this tech, it does not have to go back home to momma much for sitting in the matrix chair and learning 4 kinds of marshal arts in 30 seconds (that is kinda what R does, plug the brain in and train it like someone stepping into the matrix simulation)... Or at all (eventually).
Post automatically merged:

@Jon at Radenso You have a lot of education ahead of you to do. May I suggest you start using "buzz word" people understand like fingerprint, voice recognition, signal identifier because deconvolution is not something many understands.
Not so much for those people brain that have their AI side turned up, and their stick in the mud portion turned off :). For those chemical brain AI running full tilt, this is easy to get...
 
Last edited:

RobinAlbany

PSL +5
Intermediate User
Joined
Dec 8, 2013
Messages
177
Reaction score
185
Only device out there I can think of that does edge inferencing is Tesla autopilot. Definitely in a vastly different price range.
Excellent example! I think that's why we made the distinction by saying edge inference on a "consumer electronic device". Which is at a much lower price range, as you pointed out.
 
Last edited:

davidc502

Avoiding Speeding Tickets
Beginner User
Joined
Feb 8, 2017
Messages
575
Reaction score
710
Location
Tennessee
It is/will be used in Cyber Security.
Post automatically merged:
As I'm sure everyone expects, AI in Cyber is super expensive. I worked with IBM on a quote for some "Watson time", and I laughed in their face when it was in my hands. I'm sure it has value for fortune 500 Companies who can afford it, otherwise they wouldn't be high balling the prices. As other companies introduce AI and create competition the prices will go down.

Back on topic, I'm going to take a shot at the introductory price of said new Radenso AI (Rai) radar detector. I'm thinking $1099.00 will be the price. If this new detector blows the doors off anything out there, so am I thinking this may be too low of a guess? If so, what is your detector ceiling? I'm thinking the max I'm willing to go is 1500.00. However, if this thing does significantly crush the competition in sensitivity, then I might throw out a few more bucks. I say 1500, but the wife might have a cow and curtail that a bit :)
 

G37X Jockey

Advanced User
Premium Member
Joined
Jul 22, 2012
Messages
1,339
Reaction score
1,182
Location
Behind my Steering Wheel
Let’s talk about potential pitfall using this technology: if a signal signature has not been learned by the neural network, it will not alert therefore could potentially put at risk the speeding drivers So it becomes paramount for Radenso to update their neural network deconvolution frequently as new equipment get released.
We're already in this situation with laser jammers. We're reliant on updates to keep up with new VPR gun algorithms.
 

DrHow

Going “Plaid” ASAP (Tesla S) RDT refugee
Advanced User
Lifetime Premium Member
Joined
May 18, 2018
Messages
2,543
Reaction score
4,031
As I'm sure everyone expects, AI in Cyber is super expensive. I worked with IBM on a quote for some "Watson time", and I laughed in their face when it was in my hands. I'm sure it has value for fortune 500 Companies who can afford it, otherwise they wouldn't be high balling the prices. As other companies introduce AI and create competition the prices will go down.

Back on topic, I'm going to take a shot at the introductory price of said new Radenso AI (Rai) radar detector. I'm thinking $1099.00 will be the price. If this new detector blows the doors off anything out there, so am I thinking this may be too low of a guess? If so, what is your detector ceiling? I'm thinking the max I'm willing to go is 1500.00. However, if this thing does significantly crush the competition in sensitivity, then I might throw out a few more bucks. I say 1500, but the wife might have a cow and curtail that a bit :)
Watson is too slow. Most cloud based systems at this point are when they have to reach out to other locations back from the cloud. Yes it is expensive for now as a service compared to other options. Only are most serious clients I have find the budget to deploy any of current AI. And does it work.

I have had bunch of people ask me why in the world would a discrete base level hardware chip maker (Broadcom) buy Symantec very recently? At first not using human chemical brain AI, it looks silly decision. BUT.... Think about it. AI for the masses dealing with end point, IOT devices driven by next gen way beyond FPGA like chips with AI code to secure billions of devices. I will not go into this now because too off topic except to say this... Radenso gets this. They are driving real AI power into small low power device. They pave the way for many others. Especially in cyber security field.

remember the various reboots of Terminator? People could easy grasp “the cloud” in the movies. A “central core” taking over. Today, this human nature to be attracted to communist central control and planning is running wild. Like “cloud computing“ and data centers. Anyway, then Terminator series did a version where in the end the SkyNet was NOT central. Instead it was everywhere. Distributed and powerful. That movie did not do very well (in spite of that blonde actress Loken walking around), in Part because many people in audience did not get this concept.

Well, in Radar detecting world, and cyber security you want to protect ingesting and egress points at the end point, and various areas of the network (and many other areas). AI will be deployed in mass throughout. Traditional CPUs, memory, backplane, IOPS traversal can’t do this efficiently.

If you break down Radenso design, the traditional off the shelf hardware is used as a gateway from older style interfaces (antenna, USB, video, whatever) to provide ingesting data, or creating output for the user. It is not doing the AI function. That is left to the dedicated chips and AI code. We will be seeing this general type of AI layout moving down the road at the end point level all over the place. Not at the “SkyNet Central Core”. Sure there will be orchestration that can be in the cloud. And also external “Matrix” plugs to jumpstart the AI will be Datacenter driven (small or large).

let’s talk “Matrix” movie theme for a moment. Radenso uses a “Matrix like plug” to train Rai at the start. Think of Neo sittting in the chair plugging into the training matrix. They download 6 types of marshal arts in a minute or less. His AI brain now is trained. But, when he fights, “in the field”. His FPGA (Field programable gateway array) like brain and code then in real time assesses the opponents moves and adapts. All while learning on its own. This movie example is one way to illustrate only a portion of that is going on with Rai.


concepts: Rai like other AI only limits today is how it gets its data. Like a brain “seeing” a truck jackknifing in front of him/her while driving. The Al Brian if trained already knows what to do and adapt to that individual emergency. However, the brain has to deal with finger, hands, feet/legs muscles, motion converted to a car‘s mechanical input. This interferes with escaping the emergency. However edge inference Tesla AI network is plugged right into the car. And it’s investing sensors (visual, radar, sonic) are also plugged right in. It reacts 5000 time faster (est), than a human ever could. Or a older mechanical method could. This is like Radenso dealing with LEO threats. It is so much faster. And it’s limits are really centered more in what tools it has to ingest the data to figure out the LEO situation. Tools like radar sensors, RF police communication systems, embedded road threats, laser sensors, whatever now or in future comes out.

Today, Radenso has advanced the sensor portion by going to a SDR (software defined radio). This helps a bunch for ingesting RF signals farther away, and passing this on faster to Rai. Also, using LNA is improvement. Don’t know the laser improvements yet. In all cases, AI/Rai is still limited by what is sending to ingest and crunch. It has so much more potential than the amazing things we see today demos.

@Vortex and @Jon at Radenso. I attempted to put a different method of education together in the comment. Please let me know if this helps, or does not help.
 
Last edited:

02tbird

PSL +5
Intermediate User
Joined
Sep 29, 2011
Messages
177
Reaction score
74
Will you be testing tractors for the AI? I know most won't need it, but those of us in farm country would appreciate the ability to filter out K band ground speed radar used in farm implements.
 

Vancity23

Advanced User
Lifetime Premium Member
Joined
Jun 11, 2017
Messages
3,903
Reaction score
4,554
Location
Vancouver, B.C.
This is what I the initial video implied. Thanks for giving it even more clarity and detail.
Post automatically merged:


EDIT:
Would be interesting story to hear how you accumulated all the BSM (and door opener) profiles.
Did you have to go out and rent a bunch of cars? Sounds like a heck of a lot of field work in addition to the IT and hardware side.
No it is the opposite, they only need to learn the radar gun signatures, everything else doesn't matter as it wont alert to those...
 

poolmon

Advanced User
Lifetime Premium Member
Joined
Oct 26, 2013
Messages
615
Reaction score
916
No it is the opposite, they only need to learn the radar gun signatures, everything else doesn't matter as it wont alert to those...
Below is a quote from Radenso's post #31:

"It was explained earlier, it is totally possible that Theia tells you that it has detected 3 Mazdas, 2 Cadillacs and one Talon police gun. It is trained on all police guns of course but also on known polluters."
 

DrHow

Going “Plaid” ASAP (Tesla S) RDT refugee
Advanced User
Lifetime Premium Member
Joined
May 18, 2018
Messages
2,543
Reaction score
4,031
Will you be testing tractors for the AI? I know most won't need it, but those of us in farm country would appreciate the ability to filter out K band ground speed radar used in farm implements.
Good point! And the harvesters, sprayers, seeders machines. They seem to all have them. Wonder if they use samd makes of BSM. But the signature would look slightly different to to power and other things (mentioned in video today).
 

Penumbrian

Learning to Drive
General User
Joined
Jul 3, 2018
Messages
39
Reaction score
61
Good point! And the harvesters, sprayers, seeders machines. They seem to all have them. Wonder if they use samd makes of BSM. But the signature would look slightly different to to power and other things (mentioned in video today).
Yes, don't forget those. Pretty common on tractors, sprayers, fertilizer spreaders in farm country, and can be detected even when passing by on the interstate or highway.
 

DrHow

Going “Plaid” ASAP (Tesla S) RDT refugee
Advanced User
Lifetime Premium Member
Joined
May 18, 2018
Messages
2,543
Reaction score
4,031
Yes, don't forget those. Pretty common on tractors, sprayers, fertilizer spreaders in farm country, and can be detected even when passing by on the interstate or highway.
I have a large multi state client that is a dealer and service center for all this stuff. IF Radenso wants, would connect them together. Radenso, PM if interested. Yes, I have picked them off along Super highways, state and county roads Off axis and long range on axis. PIA.
 

Vancity23

Advanced User
Lifetime Premium Member
Joined
Jun 11, 2017
Messages
3,903
Reaction score
4,554
Location
Vancouver, B.C.
Below is a quote from Radenso's post #31:

"It was explained earlier, it is totally possible that Theia tells you that it has detected 3 Mazdas, 2 Cadillacs and one Talon police gun. It is trained on all police guns of course but also on known polluters."
Known polluters yes but I don't think there intention is get signautres for every polluter out there...
 

poolmon

Advanced User
Lifetime Premium Member
Joined
Oct 26, 2013
Messages
615
Reaction score
916
Known polluters yes but I don't think there intention is get signautres for every polluter out there...
Yes, known polluters will be taught per their comments, as it is definitely difficult to find them all and more can be added at any time.

From their comments I believe it is their goal to continue to add over time as many black and white list emitters as they can to their supercomputer's database to increase and refine the teaching; especially in those instances where there the AI's ultimate interpretation may be relegated to probabilities because their still exists some unknowns with a specific signal's profile being analyzed.

I was truly amazed at how much data is going into this process and how it all resides on their super computer and how only the resultant learning from all that data needs be passed down to the RD. It would be interesting to learn more about the resultant learning on the RD itself vs the raw data.
 
Last edited:

Heywood

Learning Something New, Still Dying Stupid
Advanced User
Lifetime Premium Member
Joined
May 12, 2016
Messages
3,749
Reaction score
6,852
Location
Alberta Canada
Known polluters yes but I don't think there intention is get signautres for every polluter out there...
The best thing is they’re looking to give you the option to see it if you want.

Tailor make it to how you want it to alert or not. If you just want to drive around and have it show you known door openers.... you can.
 

Discord Server

Latest threads

Forum statistics

Threads
79,080
Messages
1,204,220
Members
20,077
Latest member
Dallara
Top