118: Machine Learning or Machine Failing?

If the machines have learned anything, it’s that we’re talking about them and machine learning’s impact on our future.

Watch the Undecided with Matt Ferrell episode, “How AI Could Solve Our Renewable Energy Problem”: https://youtu.be/HAdiVIitI9M?list=PLnTSM-ORSgi5LVxHfWfQE6-Y_HnK-sgXS

YouTube version of the podcast: https://www.youtube.com/stilltbdpodcast

Get in touch: https://undecidedmf.com/podcast-feedback

Support the show: https://pod.fan/still-to-be-determined

Follow us on Twitter: @stilltbdfm @byseanferrell @mattferrell or @undecidedmf

Undecided with Matt Ferrell: https://www.youtube.com/undecidedmf

★ Support this podcast ★

On Today’s episode of still to be determined. We’re going to talk about whether or not the machines will listen to this and then cancel us if nobody ever sees this video then we know that they are in fact, learning hey everybody. As usual I’m Sean Farrell I write sci-fi I write stuff for kids and I’m all around inquisitive around tech and lucky for me that my younger brother was born 3 years after me and said guess what I know about the tech that’s right Matt Ferrell from undecided with Matt Ferrell is my brother Matt how you doing I’m doing. Okay yes here in the East Coast we are finally looking at spring and it looks like spring sprang right over.
I’m doing good how about you enjoying this enjoying the heat from this ah weekend.
To summer so we are looking at suddenly much higher temperatures like 25° higher here in the New York city area which means I step outside immediately go gasp I don’t know how to breathe this air anymore. Don’t know what it’s like in your neck of the woods. But I do know that. Where our parents live up in Rochester which is usually much cooler. They almost hit 90 the other day. So

today we’re going to be talking about Matt’s most recent episode which was how Ai could solve our renewable energy problems. This episode dropped on May Tenth Twenty Twenty two and right out of the gate you had sort of an unspoken in this entire episode which was there’s Ai learning which. Is clearly a benefit. There is a Ai learning which is intrusive and then there’s some Ai learning which is somewhere in the middle you didn’t really tackle that as your topic of discussion.

That’s what I would like to tackle in our conversation more than the nitty-gritty of how does this work. What is going on your video covers all of that beautifully and so I wanted to take more of a meta approach to the discussion.

And a discussion along the lines of where and how do we begin to feel like we want to draw the lines because around machine learning and where it is a you know I do I do fully believe that there is a positive to machine learning in the vein of.
Around machine learning and how we are. Okay, yeah yeah.
Self-driving cars should be able to on the fly be able to make literally split second adjustments based on the fact that they recognize certain environmental factors that could impact safety and driving. I think that that I can’t imagine anybody would say no I’d prefer that my self-driving car just be running on an algorithm that never incorporates any kind of learning aspect that to me sounds like it would be a ridiculous thing to do or you mention weather patterns. Yeah, the ability of Ai learning to become a more and more sophisticated predictive machine around what weather will be doing or yeah, all of these different impacts on big picture global global adjustments.
Right? right.
And even down onto some of the footage that you used in your video spoke of things like Ai learning along the lines of shipping and like when to stock shelves and when to prepare. Yeah, if a restaurant is doing something when is it time to start indicating like oh start putting in certain orders in preparation of an influx of customers or you know it’s a certain time of year and the ai reminds the store manager it was this time of year that suddenly there was a rush for a certain product.
Um, right.
Now is a good time to stock up I can see stuff like that as being like that’s great that that improves the user experience and the customer or just public experience of the thing and the other end of the spectrum is Ai learning to suspect certain people at the airport.
Possibly being pulled out of the crowd and having their luggage searched or Ai figuring out that I like to get a cup of coffee before I walk into the office. So maybe I suddenly get a push notification in my phone saying oh Starbucks on the corner is having a deal right now. Maybe you want to stop in.
That’s yes.
Stuff that is maybe a little bit more like somebody looking in your pocket as opposed to something that is De-Personalized I guess is what the the word I’m looking for something that isn’t about me in particular. Starts to feel like if it’s about me in particular do I want that in my life. Do I want that in my day so I guess what I’m asking you is where do you see that line in your life. Where do you see I know that you talked in your video about power usage.
Oh man. Yes.
And you have devices that allow your system to say to you hey you’ve got something going on here and that is not about tracking your power usage in a form of curtailing or benefiting anybody but you but is there a line that ah.
Company might try to push that that you would feel uncomfortable with where is that line for you.
To expand on what you were just laying out though I Want to add one wrinkle for this conversation machine learning is also going to impact jobs in a pretty profound way. It’s coming for. Everybody’s job like no job is safe like.
My job as a video creator is not safe because there’s Ai that is generating their own video content writing your job. There’s a service called Jasper Ai that can write you a blog post on whatever topic you want and I’ve been trying it out and it’s.
Yeah, yeah.
Shockingly Good. It’s disturbing So It’s like it’s machine learning is coming for everybody’s jobs for for me when looking at this like what what where I kind of draw the line from a personal point of view is like what you were laying out if it’s benefiting me directly I. I’m willing to have that kind of like give and take with a machine learning algorithm or service where it starts to get into that squishy. Okay, this is getting uncomfortable territory it starts where it starts to invade your privacy a bit where it’s like okay my Ecobee Thermostat. It’s using machine learning to understand when to. How to dial my temperature to reduce my energy use to become a little more green but keep me comfortable in my house. Ah, but they can start to ah understand when I’m home and I’m not home and they can start to predict when I’m not going to be around when I am around they can start to do all this stuff where it starts to become that tradeoff starts to become a little squishy and a little weird.
Yeah, yeah.
So there’s this tension. That’s always there with this stuff for for me around the stuff that I currently have in my house big. One would be voice assistants like when you’re calling out to a voice assistant. It’s learning from like Google’s learns from how you’re interacting with it. So it’s like the more I’ve interacted with mine. Go out in the kitchen in the morning I have a Google nest hub and it’s got a screen I walk out there. It’s like hi Matt and it shows me stuff on my screen that I might want to interact with and it’s learning what I want to interact with based on what I’ve interacted with it in the past and questions I’ve asked it and things I’ve done on Google and it starts to feel a little big brother. And so it starts to get a little like in that 1984 squishy territory of like oh do I really want this in my life but it’s adding value. That’s the thing it’s like it’s it’s that tradeoff it’s like am I getting actual value out of this thing am I am I willing to do this tradeoff.
Am I willing to give up a little bit of that privacy a little bit of that. That’s that side of it.
Do you think it’s adding value because it’s cute. Do you think that? do you think the cute is making it easier to swallow because like the little games that people play with Alexa ah you know ah Alexa Play farts for me alexa you know whisper.
Um, in in in terms of things like Alexa. Yeah.
Alexa tell me a joke. Um, and you know the machine learning going on there. 1 of the things that I’ve heard people say is well it doesn’t engage with you unless you engage with it and I call shenanigans because it’s always already listening.
Ah, yeah, yeah.
Otherwise it won’t know when you engage with it and there’s there for me is a blurry line that I don’t know that I’m comfortable with the idea that the thing is always already listening and I don’t know if you’ve had this experience but my. Girlfriend and I have an ongoing conversation around the fact that Instagram always seems to know what we’ve been talking about.
Oh don’t don’t go into conspiracy tinfoil Hat territory because I’ve.
It starts to feel creepy when it’s like we talk about a specific specific show. We’ve never watched and then Instagram is like hey have you thought about this show. It starts to get weird and that’s and and I do understand that it is like you said tinfoil hat but it is it.
Okay, so here’s here’s here’s here’s yes, it gets we.
Unavoidable when this kind of learning is becoming ever presents it starts to then trip over the human tendency to see coincidences as plans and so I think that there is an aspect of this that is.
Blurry line is already something that certain people myself included are going to say ooo no no, that’s that feels weird.
We’re kind of going off tangent here. But it’s like just for the voice assistants. They’re not eavesdropping on you 24 7 and shuttling that stuff to alexa headquarters or Google headquarters That’s not how they work the reason that services know so much about like How did you know I was talking to my friend about going camping this weekend. This is weird that I’m seeing advertising for camping the amount of metadata that’s available to all these services and then they can use machine learning and algorithms to scrape that data and create profiles for us doesn’t need. They don’t even need to tap into your microphone to to eavesdrop. They don’t have to even do that because they can create shockingly accurate profiles because our phones are telling them where we are at all times and they know oh hey Sean’s phone was near by his friend’s phone for about 45 minutes in this cafe. So it’s he’s probably friends with this other person and we know that person. Is going camping a lot. So maybe he wants to go camping so they’ll start to show you camping ads and then you’re like you were listening to me because we talked about camping. It’s like it doesn’t work through eavesdropping and this is what makes it so creepy is that the machines can collate so much more data and find patterns and things.
Right? right.
That we as humans just cannot process in the same way so that it’s it’s it’s it’s the video that my video I focused on how like it can be applied for wind turbines and finding the optimal way to operate those turbines to avoid the wake effect and find the best locations for it because it’s it’s. Collating all these different permutations that we as humans can’t do in what we’re talking about from a personal point of View. It’s the Same. It’s the same thing So There’s this this this line that we’re talking about that can be crossed very easily.
Right. Yes, and for me yeah, and for me what you just described of yeah they know where our phones are so they know we go to a coffee shop for 45 minutes and that guy’s an avid camper. So Maybe I’m an avid camper as Well. That to me is not all that different from if my phone was actually listening to me.
Um, correct I don’t disagree with that There’s a huge privacy thing where it’s like I feel like we’re going off on a tangent around privacy around this. But for me machine learning just in general is I worked in software development for a couple decades and I’ve designed user experiences with machine learning as. Part of the system to help make the user experience better. So from that point of view that tradeoff we’re talking about I’ve I’ve actually designed experiences that use that and it’s it can be incredibly powerful and then on the flip side you’ve got like I was saying like there’s. Ai and machine learning has been used shown that it can do a better job diagnosing cancer from Radiology X-rays and exams than human doctors can like it’s not a small margin. It’s a major margin. It’s better at. Self-driving. It’s better at diagnosing cancer. It’s gonna be better than writing than we are. It’s gonna be better at producing videos than I am it’s gonna be It’s gonna do everything better than what we can do so that that that line is something we’re all going to have to grapple with. And I think we need to have a public conversation as to like where we want to draw that line. What the ramifications we’re comfortable with because it’s going to impact is going to impact our lives in a very profound good way and it’s also going to impact our lives in a very profound horrible disturbing way.
Um, right.
It does leave open the question. What does that leave for us and what does that leave what does that leave and and related to that What does that do to a society that is built around measuring the value of people.
Well, that’s where I was talking.
Yes, this is this is where I come like to tire our other podcast trek in time where Sean and I talk about star trek in chronological order. Ah this is to me where it’s like.
Through the types of work they can do.
We can start to see glimmers of a star Trek future for humanity if it goes in a good direction because imagine autonomy robotics and machine learning taking the mundane stuff. We don’t want to deal with away.
And then it allows us to focus on the things that matter to us most and it completely changes the dynamic of what an economy is what it means to make a living It’s like all of that stuff is going to have to shift. It’s going to be a paradigm shift and it’s going to be an ugly transition but it feels like you can kind of see like a glimmer and oh star trek they figured it out. No currency that I want to be a I want to go into starfleet. It’s like okay, there’s no money you’re just I’m on starfleeet now. It’s like what do you want to do I want to go grapes in France I mean it’s like you basically do what you want to do.
Yeah I hope I Yeah the the I mean the idealism that you’re just talking about is would be magnificent I think that we both we both recognize that we are.
Yes, yes, it would.
Potentially right now in a hard paradigm shift transition period that nobody would be able to bet money on where this is going to end up where where these changes are going to lead and that yeah and that’s where it starts to get that’s where it starts to get scary.
It could go horribly wrong.
So to pull it back more directly to your video the systems that are in place right now. How cutting edge do they have to be in order to do all the things that you see them trying to do are we already there with. Systems already up to the task or is there still growing room for them to say like oh we’re very close to being able to do predictive weather that is accurate to a much higher percentage within a like a 3 or 4 day window whereas it used to be a 6 hour window or are we already there with like the edge of the technology being in a place where the technologists say I don’t know how much more juice we could squeeze out of this. This is really so top notch that I don’t know how we would improve it.
Ah, from what that’s good question from what I was seeing when we were pulling this together. There’s some aspects where it feels like there’s not a huge amount of room for improvement but it comes around like the how do you opt here’s a wind turbine technology. How do we optimize its placement and its use to make the most energy it can make I think we’ve kind of it’s it’s already happening today. Those models already exist. It’s already being applied today. So I don’t know how you could make that turbine better just with some more modifications and how it’s run. Think the the room for improvement is here’s a new wind turbine design that was helped that was partially designed by machines that helped us make a machine that was just even better than what we designed already. It’s like I think that’s where the huge potential is still for the future. But.
Um, right.
Right. Right.
For like existing infrastructure and taking machine learning applying it to what we currently have to make it better that stuff’s being applied every single day to what we already have so that’s that’s already here.
And how accessible is it. You mentioned Jasper as as one of those things where it’s like like it’s a website you know somebody could go to the website and say I needed an essay. Um, but how accessible are some of these things like. If somebody wanted to say oh you’re building a house right now if you were curious about how is my house going to impact you know be impacted by weather patterns like what direction can I anticipate storms coming from in my new home and and. How might I be impacted if I put trees in a certain location. Maybe that’s a bad place for those trees to be because it’s going to be impacted in this way. How accessible is that to you as a as a just a Joe on the street as opposed to somebody who is running a major company that is looking at this.
It’s not very accessible. It’s like you have to be a software developer or be working in a company to really make the most of this stuff. There’s a lot of Cloud kind of ah computing that you can tap into to do a lot of this stuff. But again you have to know what you’re doing to be able to.
To apply it and you also have to have the massive data sets that you need to be able to train the machine to do what you needed to do and so it’s like for what you’re asking. There’s really nothing that I personally could tap into to do that.
Right? And is there any information that you have that you came across that was in regards to any kind of policies or governmental regulation around this I know that there was a point where Google decided to turn off machine learning in its Google translate. Operation because what they discovered was that Google translates learning algorithm had created its own language because one of the things it had learned was sometimes a direct 1 to 1 translation was not the fastest route to translation. Google translates learning algorithm figured out that oh sometimes it’s easier to take the translation into this own generated language a basically a second language that was a hodgepodge of different languages to then get to the final destination. So maybe going from german. Japanese wasn’t a smooth translation but german through english to japanese was better and that learning algorithm was creating its own hybridized language and the scientists at Google decided to turn it off because they didn’t know what it was doing. They were not quite sure whether or not. It was in fact, an artificial intelligence. So my question to you is? are you aware of anything and spoiler alert I’m going to say there’s probably not but is there any action on a governmental side to say let’s really keep an eye on how all of this is going and how it’s being utilized. And what’s going on with either invasion of privacy or accidentally creating something that’s going to wake up one day and say I don’t like what’s going on I’m going to shut down the power grid.
There’s on that second point, there’s nothing. There’s nothing. No, it’s it’s it’s it’s a concern from a lot of people that that’s that we need to get rules in place. But then the questioning comes of who makes up the rules and how to be agreed to it.
Yeah, yeah, I’m not surprised.
Because it’s like we could come up with our own rules for the United States but then in China they could have completely different set of rules and it’s no holds barred and they’re coming up with Ai that’s going to destroy the world and we have no way to to say hey could you? ah please stop doing that. So like there’s there’s there’s there’s nothing for that. Um, yeah, but for the first.
Um, yeah.
For the first point you brought up around like privacy and things like that some countries and regions like the eu for example has really tight privacy rules for computers and online activity. So it’s like there are areas that of the world that have guidelines around how this data can be used and. To create models and how it can be and how we as people have a ah right to have that information taken away and given back to us so it varies depending on where you live for that stuff.
So that’s kind of our way of wrapping this all together into a big bundle that says Well, there’s still a lot to be determined. It’s nothing has really been decided upon I think that this this is one of those conversations though where.
Yes, there is.
Very often Matt and I come into these topics and we’re both like wow self-driving cars that sounds cool and this is one of those topics where we both show up and I’m on the opposite side of the fence from him and I’m saying I’m a little creeped out. Why am I worried and he is like hey future.
Yeah, what more my response is you should be worried but at the same time. It’s kind of awesome.
Maybe it’s not so bad.
Right? So I’m curious to the listeners. How do you all feel about this please do let us know you can reach out through the podcast description where you can find the contact information or if you’re on Youtube you can just scroll beneath the video and you can find. Comment section and leave a note for us there while you’re doing all of that you can leave a review you can like and you can subscribe all of that really does help the channel and if you want to more directly support us. You can go to stilltbd fm. Click on the become a supporter button and you can throw some coins at our head. You can also throw coins at our head in Youtube by going to the join button and becoming a member there. All of that really does help the channel and we thank you so much for listening or watching we’ll see you in the next one.

← Older
Newer →