How'd you like to listen to dot NetRocks with no ads? Easy? Become a patron For just five dollars a month, you get access to a private RSS feed where all the shows have no ads. Twenty dollars a month will get you that and a special dot NetRocks patron mug. Sign up now at Patreon dot dot NetRocks dot com. Hey, Carl and Richard here with your twenty twenty four NDC schedule. Will be at as many NDC conferences as possible
this year, and you should consider attending no matter what. The Copenhagen Developers Festival happens August twenty sixth through the thirtieth. Early bird discount ends April twenty sixth. Tickets at Cphdevfest dot com. Ndcporto is happening October fourteenth through the eighteenth. The early bird discount ends June fourteenth. Tickets at Ndcporto dot com. And we'll see you there, we hope. Hey, welcome back to
dot net rocks. I'm Carl Franklin and I'm Richard Campbell. We are past the nineteen hundred mark here post build yeah yeah, we're also this is like the one studio show. We're doing in between a whole bunch of in person shows, right right, So it's gonna be fun. Yeah, we're going to be at NDC Oslo very soon. And yeah, really good feedback from show nineteen hundred was Scott hansomen people really love that we had some fun there,
you know. And that was your idea too, to yeah, a surprise him with the with the quiz the clips and he got one hundred percent. It's ridiculous, right, I know that's sky. All right, let's roll the crazy music. All right, man? What you got? Well for bed, I know a framework today I don't have anything about the framework, but okay, but on my other show Security this week, we found a hack of vulnerability in a TP link gaming router nice that exposes users to
remote code attacks. If you look at this thing, I'm thinking. The first thing I thought was, what's the difference between a gaming router and a line of business router? And the gaming router has more blinky lights and it costs more, yeah, and cool colors and things too. But yeah, it also prioritizes with QoS, It prioritizes the game packets, and they tend to use as many channels as they can you pay a premium for these things?
Yeah, exactly. But it also kind of looks like it could take off, you know, like it's a like a drone of some kind. Yeah, the idea that your router is a statement piece for your coffee table. I don't know. It's like nine prongs sticking up out of it with red stripes on it. Yeah, they're in tennis. Yeah yeah, all
right. So people who know security cvees, which are you know, threats or reports of problems, know that there's a CVSS score with them from one to ten, right, and ten is oh my god, this is horrible. It's the worst possible thing, right, but it only that score is only if you have been attacked. How bad it is opposed to this router? I presume, well, yeah, if somebody has attacked the router and hacked it, right, it's bad, like remote code execution is bad.
But we at security this week have started coming up with how likely are we to be exposed to this number? So it's called like a contagion factor or something like that, which is not part of this number. So we went through the problems and first of all, you have to be rebooting. It has to have just booted up. So if your power goes off, your power comes back on, or you turn it off and turn it back on
right in that boot up sequence, that's where you're vulnerable. And then certain ports have to be open right right, and then you're vulnerable to this attack. And my question was, well, does the attack have to happen from inside the firewall or also outside? And the answer is outside. It's exposed to the Internet, which makes you vulnerable in those seconds when it's booting up.
And if somebody knows about that and sees it and connects to it and sends you the special blow up commands, they will get remote code access. Right, It's pretty scary. Yeah, it just means that. And you've got to know there's some script kiddies running this, running that call continuously, just walking back and forth. It's IP addresses, right, Yeah, that's right. So anyway, this is a kind of a big deal, and
I wanted to share it with people. So, yeah, you know what, patch your routers, patch everything regularly, patch all the things on run as a little while back into a show with Sammy Laho is one of these one of the security folks. It's scarce it's not out of me on a regular basis. And and he talked about the fact that with multi factor authentication and you know, current security password technology, phishing is stopped being the primary
exploit PLATH. Unpatched gear is now the primary exploit PLATH. So we had a whole conversation about you know, most IT folks want to test before they up they update, and so often it takes a week or two before you get updates onto your machines. It's like, that's now a higher risk than the risk of an outage by applying the patch immediately and then figuring out if it's a problem. Yeah, and I know that we're going to be talking
to Veronica about AI. But AI is now in the game of hacking, you know, searching for vulnerable as soon as they're reported and immediately going out and trying to exploit them, generate and exploit, so generated exploit. It speaks to how we've got to release this information more carefully too, Yes, like more time to patch. But yeah, it's an interesting time, isn't It's an interesting time. It's you know, it's cat and mouse. But
the sooner you patch, the safe you are. Well, it's also we've got there are professionals now, like these folks are not hacking for fun. They've hacked for money, and then so make sure they don't make any money, right, all right, who's talking to us today? Richard grabbed a comment off of show eighteen seventy four, which we do backup November twenty three
with our friend vishwas talking about his applications of large language models. Vishwats is always good for that sort of thing, right, Yeah, when you know you think about We recorded that show early November twenty three, so he already had code out in the field. And that's what I love about vishwas it's like he calls me when he has something real to show. He's not speculating. It's like, let me talking about this project we have out in the
field. And this was an app that he that he was working with where the large language model was understanding texts coming in from various sources. It was almost a kind of send him an analyzer, but more sophisticated by far. And we've got a hilarious comment here that I had to read from a zirene. I could presume it's an acronym or a pseudonym. Just a few months ago. And actually it's just quoting me. All he says is you don't ask a stochastic parrot for facts. That's not what it's for. Yeah,
that's pretty good stochastic parent. Pretty easy for me to agree with myself. So siron, thank you so much for your comment, and a copy of music code By is on its way to un If you'd like a copy of music code By, I read a comment on the website at dot rocks dot com or on the facebooks. We publish every show there, and if you comment there and I read on the show, we'll send you a copy of
music code By. And also you can follow us on Twitter. We've been on their ore ex Twitter as I'm calling you now, and we've been there forever. But the cool kids are hanging out on masthadon. I'm at Carl Franklin at tech hub dot social, and I'm Rich Campbell at Macedon dot social. Yeah, send us a two. Those are other ways that you could
get yourself a free copy of music to code buy. All right, well, let's bring on our esteemed guest software engineer by day, an AI enthusiasts by night, Veronica Kolesnikova has extensive experience in full stack development using c sharp, dot Net, Java and Typescript, Azure and AWS clouds. Public speaker, hackathon volunteer, co organizer of Boston Azure user group, and mentor. She's curious about mobile and augmented reality Veronical's a master's degree in information technology.
She also likes dancing, traveling, and practicing arial yoga in her free time. Arial yoga suspended from a helicopter, right, of course. If I had not heard of aerial yoga for forever and like in the past two months, I've thought, you're like the third person, like what's going on? Like if it's as if yoga wasn't hard enough. Now you're hanging in the air while you do it, and it's it hot or cold aeriel defense. Yeah, if I practices in Florida, then yeah, it's still hot,
but here in Massachusetts it's sometimes cold. Yeah, it's just starting to get hot now. Yeah, So what's on your mind in the world of AI? Forroonica, everything is changing so fast every day I feel like I'm behind, although I'm reading the news and trying to connect to people online, but still I feel like I'm always behind. There is something new happening all the time. Yep, I feel the same way. It's not a bad thing, you know, like really we don't get sprints like this all that often,
right with a new technology. I was explaining you to some high school students last week or earlier this week even I said, Look, the scientists come up with a new material, and then the engineers take the new material and they try stuff with it. What you're seeing right now is the science has been done. General of AI is where it's going to be, and the engineers you're running with it now, and we're just going to make some stuff and some of that stuff will be bad and some of it will be
pretty useful. Yeah. A lot happened actually recently with Jenny AI and large language models, and now I feel like there is another trend coming with small language models. Yeah, how do you define the difference between its large language and all a small language model? The science? Can you put numbers around that? Is that like a number of parameters? Is it actually like the physical data size? Yeah? The data size is the major factor. There.
Lots of those small language models you can actually put to a mobile device. So if you're created a mobile application it can run on the device, which is really cool. That seems to be the goal. Is I want to run this on a phone? That's debatable. Yeah, want is the struggle word, Like, I don't know that I want to run this on the phone, But it seems to be what they're offering us. Isn't the context of a small language model usually smaller and more focused on a particular area.
Is that the story? Like, you wouldn't ask a small language model the current price of a foot long tuna subway, it would be right, Yes, yes, exactly. So large language models they were trained on more data. That's why they're large, and they're more they know more about stuff, and then small language model us they're more focused on specific areas. Yeah. Well, I don't think i'd ask any of them what the price of a sub sandwhich was because that's a fact that they're not good at facts.
Yeah, they're not, are they? Yeah, you have to use RAG for that. Yeah, I'm sorry. We've got to be accurate in police RAG is which we will augmented generation. There you go. It's such an as. Yeah, the terms we have to live with now or did you did you see any of the keynotes or the sessions from build about Microsoft announcements around their AI stuff. Yeah, I was following online. Unfortunately I couldn't make it to Seattle this time around, but they released a lot of stuff.
I'm glad that they're focusing not only on being up to date and up to speed, they're partnering a lot with open Ai, and I think it took them basically a week to include GPT four oh into their offerings as well for as open ai. But also they're focusing on responsible AI, which is very important. That impressed me. I mean, I don't know how responsible or secure it is, but the fact that they have this security layer between you and the models and the outputs and inputs makes me feel asleep a little
bit better at night. Yeah. Yeah, And I think there was a really issue way that they presented it with the AI studio, where you know, one hand, it was like we're here to protect you, but it's also there's also a sense of we're here to also protect people from you too.
Write like I don't know how a bad actor could use Microsoft's AI tools and not have the tool essentially stop them saying this is not an appropriate use, and that that to me, you know, relieve me, because we've been you know, we talk about open AI and this whole idea of an open model. It's like, do we really want these tools out in the while where anybody can do anything with them? Yeah, same as open source
projects. So everybody here is probably old enough to remember the heyday of SQL injection attacks, which, by the way, are still the number one vulnerability in software. SQL injection attacks number one. So if you think about what a SQL injection attack is, you know, you have a field on a web form that you're filling in, and because you're smart and you know what
the database is, you can embed SQL commands right in that data. And assuming that the back end code just concatenates all those inputs together without sanitizing them or without checking them, and sends it as a query to the database, therein lies the problem. Right. So think about in the context of AI. You know, the even the APIs like the open AI APIs take a sentence, it's text, right, So now we have prompt injection attacks.
It's like a whole new level of threat. So they have to be a very diligent about sanitizing and discovering ways that people are trying to either hack into the system remote code execution or or you know, injecting and getting into the data model or something like that, or just asking stupid questions or dangerous questions. So it's really good that they have this this filter. Yes, people come in with creative ways to break the system, and now Microsoft is that
in the jail jail breake detection, which is very good. But you know, it's always cat and mouse. Who is more creative, people who created those jail breaks or people who are trying to catch them. Yeah, and yeah, I tend to bat on the white hats just because there's a lot of them. When they're pretty skilled and they don't have to hide. They get to share each other's information and collaborate with each other, where the bad guys have to stay in the shadows. All right, Well enough doom and
gloom. Let's talk about some positive things in the changes in open AI, maybe in particular or AI in general. What do you see out there? Yeah, so they just released that new model GPT four oh, which looks like it can do everything. It can analyze visual content, speech and text all from one place based. I remember the days when we had to use different APIs in order to process all kinds of information. But now it's not the case, which is very convenient. Those days were like a few months
ago. And this is what they mean when they say multimodal, right, that it doesn't matter whether it's text or images or it can pandle it. Yes, okay, how are you? I mean four oh seems niche also in the sense that it's kind of a client. I don't know how it embeds in my apps. Yeah, you can create all kinds of applications with
it, all kinds of assistance or I don't know, support tool. I know some people they are struggling with creation of graphs or maybe processing information from graphs, and by just connecting GPT four oh to application like that, it's easy for people to maybe read the graphs, maybe create new ones based on the data that they have. That's the use case that I was thinking about right away, because I remember some of flower and we are talking about like
the Excel style graphs like graphs and overload a term these days too. Yeah, yes, interesting, So take up high chart. By the way, never use a pie chart. You send it to four oh and ask it to analyze that, Like, what do you get from this, what would be a better way to represent this data? Because that's a great problem. Yeah, I'm sure you can do that. And I know that some of our project managers they have to create all kinds of graphs pretty much every day,
and they have to process a lot of data. But now they have that really good support from GPT four. Oh interesting. Yeah, I've been looking at these tools for a while. Now is good summarizers, you know, and you know, it's funny to think in terms of graphs are supposed to make it easier to read data. But I'm in the same situation as
a lot of those managers. I get twenty pages of visualizations of information where I'm like, how do I summarize, you know, get this down to one thing, you know, to be able to push it through a tool to say what's the one page I should look at, or you know, what's the sum of all of that and still have it being meaningful. That's an interesting problem because I'd have never had a tool that could actually take a bunch of graphs and give you a single description. Now, have you taken
this out for a spin? Like, have you tried doing that? I tried to play with it a little bit. Unfortunately, didn't have enough time to actually test it really well, but small scenarios something that they showed at the presentation that open Ai did a couple of weeks ago that worked. Look like it worked. I'm sure there are some downsides too, like I saw people complaining online that it's a little slower than GPT four and GPT four vision. But you know, there are always ups and downs. It also was
brand new, so everybody's playing with it at the same time. I'm sure it's under a lot of load. Yeah, so I took it for a spin. Of course. I have chatchapt four oh in a browser, and I don't see any speech or anything, you know. I know, I've used my voice to communicate with chat GPT four on my iPhone version before four oh came out, and I could even and you can even attach pictures and
upload them. So I don't really know. And I don't see any like speech things or anything different in the interface on the web anyway for chatchipt four oh than I did for four So how does that work? Yeah, so I think for end users the difference is maybe unnoticeable, as you said, but behind the scene, I think that they just had to convert different inputs into texts and then pass it to the model. So the API has endpoints
now that accept streams of audio and video, et cetera. Yeah, so in theory it should be faster because we are eliminating that conversion part of it, right, Okay, Yeah, I'll have to check my phone version. But I didn't see anything in that either that was any different. Yeah. I guess this is about digging into the API enough to see how the different
sources could be fed to it and and then what it will return. I mean, I've always had it return as checks right, like summarize this additional information, but I've never tried to have it returns like draw me a better graph like that, because that also seems very precise for a tool that tends
its outputs don't tend to be that precise. So on the in the mobile app, which I updated so it supports four Oh, now you have you know a camera, you have you know, select a file, you have select an image, you have a microphone, and you have headphones, which I'm thinking is going to be you know, speech output. So so that's good. The web version, however, no different. And I couldn't find anything in the settings to enable it either. So then, yeah, did
my homework limitation on the web Pageah, we'll see. I even asked chat GPT, right, how can I use speech to interact with chat GPT four? Oh? And it says to use speech to interact with GPT four, Oh, you need to utilize voice input with voice recognition or voice input on web browsers like Google Chrome or Speech to Text APIs right, So it really doesn't offer any anything at the at the app level. It sure sounds like Microsoft documentation too, right, Like, yeah, it's an answer but not
actually used, not very helpful. Maybe I'm missing something. If I am, somebody email me please let me know how this works. I yeah, obviously all the demos I've seen, it's got that her vibe you know, the movie and even to the point where Scarlett Johanson I believe, is suing and I'm like, ah, why are we doing this? Why are we going down this path? I know it's so stupid. I'm trying to get this out of science people not think about this as science fiction and worry about
the fact not helping. Yeah, they're really not helping. Actually. Sam Altman posted on x Twitter, saying her a couple of days out they release Yeah, you know it's a copyrighted term there, Sam, right, like what are you doing? Yeah? Yeah, it's not Scarlet's voice, but it's someone who sounds like Scarlet's voice. Yeah, it's very very close. And then again it's like, why would you do that? Scar Joe is
kind of famous for being willing to take on anyone. She took on Disney for crying out loud and one yeah, and then still gets to work there, Like who does that? Anyway? I Yeah, I wish they would stay away from the science fiction parts, just because I think it's bad for the for people to get caught up in the non facts. That mean, it's good for fundraising just because Okay, if you want to deceive investors, great, but trying to deceive regular people it's not not appropriate. Hey,
we should break for a moment for this very important message. Hey there, this is Jeff Fritz, the purple blazer guy from Microsoft, letting you in on a little secret about my friend Carl Franklin. You know, the guy who started dot net Rocks, the first podcast about dot net, way back in two thousand and two. The guy who's been teaching Blazer on YouTube since
twenty twenty. Yeah, that Carl Franklin. Well, Carl has joined up with Codina Castle to teach a week long hands on Blazer class this September. And get this, it's at a castle slash villa in Tuscany. It's sort of a luxury vacation with Blazer learning built in. He's calling it the Blazer master Class. You'll learn Blazer from the ground up, finishing the week with the ability to build and deploy Blazer applications. Now, Carl did this class
last year and it was a huge success. You can watch the awesome video from twenty twenty three on the website, which I'll share in just a few seconds. Since the training happens for only four hours in the morning over six days, you can bring your significant other with you and you should. This part of Italy is absolutely beautiful. There's so much to see and do, and Larry and Marco from Code in a Castle are organizing daily activities both at
the castle and in the area. The castle is in the Marema, a less tourist y region of Tuscany, offering both classic Tuscan hill country as well as easy access to the Etruscan Riviera with sublime local food, wine and olive oil around every corner. Breakfast is included every day. There will be two communal dinners at the castle, bookending the experience, and most other meals in all activities are included. And did I mention you'll learn Blazer in person from
Carl Franklin. Space is limited and for good reason. This is a quality training experience in a beautiful setting. Go to Code in a Castle dot com slash Blazer twenty twenty four that's b la zo R two zero two four. To take advantage of this amazing opportunity to join Carl in Tuscany for an unforgettable week of ladocha vita while advancing your programming skills in this important new technology.
You know it's common for business application to contain fifteen percent repetitive code just because of metaprogramming limitations in the C sharp language. Why write boilerplate manually when a machine could generate it for you? Enter Metalama, the code generation and Verification toolkit for C Sharp. Their C Sharp to c sharp template language is simply amazing. Logging caching memento observable if it's repetitive of Metalama can automate it.
Visit metalama dot net today and learn to automate your code patterns with their free edition. Remember it's Metalama with one L M E T A L A M A dot net. Now we're back, it's dot net rocks. I'm Richard Campbell. That's Carl franken Hey talking to our friend Veronica a bit about the new GPT four O and that's a oh, is an omni zero? Because why would that be confusing? And just a reminder, if you don't want to hear any ads, you can be a patron for five bucks a month.
You can get an ad free feed. Go to Patreon dot dot nenrocks dot com. All right, awesome, where we we were talking about the things that you think are interesting that are coming up here and that you've played around with. Is there something that you have not yet seen that holds a promise that that you think is really really valuable for people, not just developers.
That's a good question. I feel like those companies are so creative Open Ai and then Microsoft joined them, Google Italy as trying to catch up. Now we'll on mask with x Ai, which I heard a lot about but I haven't trusted it or I haven't touched it. Actually. Yeah, so there are new opportunities every day and new functionality. I feel like they're improving
of something like copilot for developers, for example. It is pretty good now, but there are still a lot of opportunities for improving it and making it better. I'm always laughing with my colleagues. We got access to COPI a couple of weeks ago, maybe a month or so ago, and I'm always fascinating how it tries to fill out like text. So you type in text name of your variable and it's trying to feel it for you, or a name of a LaMDA function for example, or connection string, and screen fills
up with gray letters. Yeah yeah, but I'm like, is it useful? I don't want it to fill out everything. Apparently you can turn that feature off if you don't want it constantly suggesting things. If you want to just read your comments and generate code, you can do that too. I find that chat chat GPT I use when I need like more lines of code to generate than co pilot is going to do very easily. Copilot's really good, I think for small algorithms, small blocks of code, but if I
need it to if I need some serious generation alice chat GPT. So but between those two, it you know, really helps a lot. I saved hours just in the last day implementing some stuff that was really really uh you know, case heavy, like switch heavy. You know, like there's there's this I don't want to go into details, but there's this class with like four hundred fields in it that represent every piece of a CSV file. And
you know that's the way they wanted it coded. And so if I need to do anything with those with all of those things, it it's a lot of a lot of repetitive code. And chat GPT made short shrift to that. It was great. How you tried the co pilot chat co pilot chat have co pilot chat. Oh so yeah, so that's right from Visual Studio. You can use that, right, Yeah, I have not and VS code works there. Yeah. What's better about that than say, using chat
GPT. It's just convenient. You don't need to jump back and forth. You can also reference your code that you've already written or specific files you wanted to reference as well. Yeah, Okay, there's so many GitHub co pilot products now because there's there's also co pilot workspaces, not to be confused with workspaces, because why would anyone get that confused. Yeah, but I just got my my beta access to it, and I haven't had time to really
take this out for a spin. But boy, talk about an interesting collaboration tool. It's just to have more information available to you and to think project wise. Like I'm with you, Carl, little chunks of code, right, like that seems to work fine. I don't think about discussing with the tool the architecture of a project. No, but you do need to you know, I find that with chatchipt anyway. You need to give it just
enough information so that it can, you know, do stuff. So if I have something that requires, you know, four hundred cases in a switch, right, I do the first ten and then there's an obvious pattern there and I'll say here fill in the rest of these, so very internal, right, let the intern do Yeah, yeah, yeah, all right, I mean I respect that that's useful. It saves time, less typos maybe right, No, it worked perfectly, did exactly what I asked it to
do. Yeah. Well, anyway, I saw an example with GPT four oh, when they just provided the diagram and then ask it to create code for that application. It wasn't like super complex application, but based on what they posted online it worked. Was it like a UML dihighground? Yeah? I think it was just simple A couple of boxes connected together and then flow chart. I saw that. I saw a Twitter post about that maybe a few months ago, but I never saw it in action, never tried it.
Interesting idea because I remember seeing that early in early days of Azure they had sort of a diagrammer you could use in the portal for you know, laying out an app and then it would generate like the network calls and and
vms and things like that. Remember Windows Workflow didn't write code like it's interesting to actually, you know, build that flow or draw out the flow and have it generate the code, because and I've seen it go the other way where it's like show me a flow chart of this code, you know, which I think is super useful too, because I hate trying flow charts and often I end up writing a bunch of code. It's like, what have I done? The fact that it could tell you that that should be a
tool in visual studio. Do you remember Windows Workflow foundation briefly I'm trying to forget, but it seemed awfully cumbersome without you know, it was a one point oh product. It came out, We're dot net three. It's probably part of the fallout of Longhorn for Crimelud, and I like the idea. They were trying to make a generalized workflow engine. It's just a that's a really hard problem be. It was tainted with the disaster that was what was
Longhorn. Yeah, yeah, we're still getting over it. But now apparently, you know, maybe you could use a tool like visio to create a diagram which includes classes and variables and stuff and just say here, build this, yeah, and build a framework for you. That's what I want. Really, I don't want the whole application. I just want it do.
I just want help getting started with stuff that's gonna be several types, you know, several class and lists and things, and you know, implement a custom list that overrides the you know, the methods for adding into leading and updating and all that stuff. You know, things like that, stuff that I really don't want to write. I like your word, Richard the intern Yeah, it's like, here, write this code. Then, yeah,
do you see these tools starting to take lead on projects. Veronica, Like, this is really a thing, because you know, we've had We've had folks talk that way too. It's like, when are we the intern and the tool is now you know leading? Yeah, I don't see right now, who knows in a couple of years. Oh, I've read too many books that say, listen, there's no determinism in this software generated I is good some stuff, but that's not it. I mean architecture as checklist.
Sure, it's like, hey, I want to lay out an architecture, and now the tool is going to keep your mindy you to fill in these bits like have you done this part? And have you done that part. That's what they're talking about with workspaces is like now that the LM is the nagging project manager saying you said you'd get it done by this time, and you said you do these pieces like it's making sure all the boxes are checked, which I got un useful because I don't mind yelling at the automated PM.
Well that's when you feel like the tool. That's Oh, I'm the tool. This code created by a tool. It still seems like a partner for the most part. But we're only a year into this for crying out loud. Yeah, but it's really helpful. I remember it took me a lot of time to actually search for something online and get answers, so maybe
I obviously don't remember all the functions ever or like libraries. Yeah, so it was taking me a long long time to find something specific, but now it's getting really easy with al Alam and Jenny Ai, whether using GPT or using GILB copilot for example. I like what Microsoft is promising with you know, your databases and all the stuff that's in your one drive and all the
stuff that's in M three sixty five. It just becomes you know, vectorized, and so you can use natural language just to find things in your email for example, or on teams, just stuff that happened that you don't have to go searching for. Sort of like the search idea has just taken over AI in terms of you know what Microsoft is doing anyway, they but we can't stop. We can't leave here today without bringing the elephant in the room,
which is the R word recall. So something that Microsoft announced that would be coming to Windows eleven soon that's going to take pictures of your screen and keep them somewhere. I don't know, maybe in the cloud and then let you and when you do these semantic searches about things that you were doing on a particular date and time, when was I talking about that? Blah blah blah, Like it's going to use that to figure things out. But it's
also you know, the the privacy nuts are going crazy over this. Yeah, except these tools existed for decades, Like there's nothing there's been there's been a widget for taking screenshots at regular intervals for like twenty years. Yeah. I don't know why this is different that people are flipping about it. Well is it? You know, if you don't like it, don't turn it on. I was under the impression it's going to be on by default, which is different, and now you have to turn it off. No,
no, no, no, okay. You can also search through your history. So you were taking screenshots, you still had to go and manually search for information or we were back to it, and now it's all automatic. It's going to be factorized, right, Yeah, I mean that's the big difference now is we have really great tools for looking through all those screenshots to actually find what you were doing. Yeah, and it is I don't know. I was confused about the amount of data they're going to store. Yeah,
okay, I just found I mean, how many screenshots? How much space do all those screenshots take? Yeah? Yeah, it's going to use local storage, so that that at least is good. You know, it's not sending stuff up to the cloud where other people can mine it for information very easier. Local storage is going to be filled up in a couple of days. Yeah, I don't know the answer. That's a good question. When or when do you let it purge, clean itself up to I mean,
there's tiers here, there's recall for your own use. We you know, as soon as you go down the M three sixty five path. And we talk about this on run as all the time. It's like, now you're in the corporate surveillance side of things, which there's some legitimate parts to that, Like you know, I've spent enough time working with email administrators where it's like, listen, if you harass someone via company email, the company's
liable. So the company has an obligation to be aware of what's going on inside of email. That does mean read everybody's email, but it does mean there are tools watching for harassment that then can lift it up to the point to say, hey, somebody else should be able to see this and do something about it. And so I'll recall as a tool around activity inside the company. That's interesting, But you would hope all this stuff is pretty tightly
locked down. But you know what, I like having a dashcam on my car, you know, because then I have an incontribusible copy of what happened in my car, how fast I was going, when when that guy did cut in front of me, like all of those sorts of things. I kind of like this dash cam for my computer too. What was it I was doing? Yeah, when that when I had that problem. If I have storage space, I'll use it. What I don't want is a dashcam
that's sending my data to somewhere. Yeah, And olers don't know why you would think that, seeing how Microsoft is supposed to be the privacy guys. Yeah, well it's not the case. So I was concerned about that, but you're right, it's not the case. And it does say pre rex are a co pilot plus PC, of which I think there's like none right now for and there's only well there's only beta versions or alpha versions. Yeah,
no, they it's a particular machine. These are once they announced at build, right, you can pre order for June, so by the time this show is out, in theory, you're gonna be a get one. But it's got a co pilot key, which is really important. But it has something else too, right, is there some sort of chippy stuff built in for AI that isn't on a copilot plus PC? Well, the ones that they were announcing it build were either Windows on ARM Snapdragon yeah right,
Snapdragon. And it's got a pretty seriously powerful MP unit like forty five tops, which is the current measurement system, which is questionable but okay, questionable meaning questionable just because like measuring the measuring the computational potential of a neural processor is an interesting problem. Like those are very academic benchmarks in their early stage benchmarks for new technology, So you know, how will we be sure that
that number is meaningful in terms of actual use? Are you gonna order or buy a one of these Snapdragon PCs Veronica? I wasn't planning to the relatively reasonably priced. I think that the mid range unit was like less than two k USD And somebody asked me if we're going to do a dot at rocks on Windows on ARM. I'm like, that's a pretty short show. Yeah, sure, yes, you know, yes, yeah, we done like that. I don't even know if you'll need to recompile. I suspect the
gitteral just go oh it's armed, let's go. Yeah, I think so too. That's funny, right, like, okay, yeah, not really think they the CLR team solved this long ago. I want to see how successful it's going to be. It's not all Microsoft products, especially on hardware site, were successful, that's right. Well, they do have a bunch of third party vendors. The asis is and Lenovo's and so forth are supposed to release machines too, and the Snapdragon ships that is in pressive truly,
that's pretty cool. But I'm with you. You know, generally speaking, the surface units are the most expensive versions of this. Like, i'd be interesting to see what an Asis or any of the other vendors actually can make from that gear, and you know how regional price would be. Battery life should be epic, right, I mean, that's the whole thing with the ARM chip set is you should be able not even need to charge every day anymore, right, which that would be nice so Veronica, what's in your
inbox? What's coming up next for you? So? I am planning to actually play with all the stuff that was just announced Build, especially on responsible AI side. There are lots of new tools and it is really interesting. They showed them I built. Looks like they're really cool, working, really great. I want to see it for my self. Speaking wise, I am going to Kansas City in a couple of weeks. That's the show we're
going to both be at. Yeah. Yeah, I won't be there, but we could probably do something remote, Richard, if you want to. That's fine. We got We're gonna have so many shows between Build and and and OSLO, and I think we're going to be well into August, so
I wouldn't too far ahead. Well that's really great. What's the talk you're doing in Kansas City, Fornica. So in Kansas City, I will be talking about responsibily I thanks to Microsoft, I probably have to change half of my talk, Yeah, because I before the talk was mostly focused on the responsibility. I inside the machine learning studio that they have and they still have it, the dashboard, then the report, but now I feel like they're
moving more towards general of AI analysis and all those content safety tools. So I do want to include more of it into my talk. And is that part of as your AI studio. Where's a lot of those safety tools living? Yeah, so I know they have a separate portal, but I prefer to access through them as your REI studio. Yeah, which makes sense. It's a really great place to work. Yeah, and certainly get to you know, you get a huge dashboard of choices, like how many models would
you like to play with? Because there's way too many. Yeah. Yeah, I'm looking forward to using that studio as well, just to see what the heck you can do in it. The demos we're impressive. Yeah, with that security approach where it's going to keep you from accidentally making a bad thing as well or putting yourself in danger or inappropriate responses like I don't know,
I got to think the average dev is pretty free. Doubt about incorporating this as software when you when you have all these terrible stories of how badly it can go off the rails, right right, You know, I came for tech support. I got told to get a divorce. We need more positive stories. Stories are boring, Carl, No, they're not they're amazing. I mean I told one today. You know, it's just but it's
just a day in the life, you know. How about you, listeners, Do you have any positive stories of interactions with AI or negative ones for that matter, but we want to hear the good news. Let us know. You know how to get in touch with this well. Veronica, thanks very much for spending the hour with us. It's been a pleasure talking to you. Thank you so much. You're welcome. Yeah, see in Kansas and we'll talk to you, dear listener next time on dot net rocks.
Dot net Rocks is brought to you by Franklin's Net and produced by Pop Studios, a full service audio, video and post production facility located physically in New London, Connecticut, and of course in the cloud online at pwop dot com. Visit our website at d O T N E T R O c k S dot com for RSS feeds, downloads, mobile apps, comments, and access to the full archives going back to show number one, recorded in September two thousand and two. And make sure you check out our sponsors. They
keep us in business. Now go write some code. See you next time. You got jam Vans