The Future of AI, Cloud Emulation, and Cybersecurity | Tech News of the Week

[01:00:00.000]
Announcer: Welcome to Tech News of the Week with your host, One Blue Shoe.

[01:00:07.260]
Ned: Welcome to Tech News of the Week. This is our weekly Tech News podcast where Chris and I talk about four things that caught our attention this week. Chris, why don't you get us started with something not AI-related, right?

[01:00:22.140]
Chris: Oh, buddy. Sport. What are you thinking? All the major AI models continue to lose money and stop advancing. Have we reached peak GPT? All signs point to definitely maybe. Multiple anonymous reports from OpenAI, Google, and Anthropic are stating that the advances in AI model development have stopped. What's the word advancing? Openai's latest release, Chat GPT 4-01, great name, has been receiving mixed reviews. Anthropic's latest Claude is, by all accounts, Debuslame.

[01:01:11.850]
Ned: Hold on.

[01:01:13.380]
Chris: And Google is also here. Out. Seriously, there's a thing making the news where somebody was using Google versus Anthropic for therapy. And first off, don't do that.

[01:01:26.690]
Ned: Yeah, stop there.

[01:01:27.380]
Chris: But second of all, Gemini's response is, We're terrifying.

[01:01:30.030]
Ned: Was it telling them to eat glue and bury themselves in the sand?

[01:01:34.670]
Chris: Well, it started by telling them to watch their tone and went downhill from there. Excellent. Now, none of this lack of progress has stopped or even slowed down the expansion of the frankly unimaginable compute power required by these behemoths. This, as the kids say, is a problem. So not only are we in a situation where there is no financial gain from these models, Remember, they all lose money. There is also declining technical improvements, model over model. Surely a worrying sign for investors. Or at least it would be if investment was based on anything rational at all anymore. As it is, we are deep into the Yolo economy, so naturally people have been continuing to throw money at these companies as fast as possible. It's going to be interesting to see what happens, though. In a recent discussion on Reddit, Sam Altman said that work on ChatGPT5 is still ongoing, which makes that work over a year old at this point. It will be interesting to see if this new model is any better in any meaningful fashion or maybe we've actually hit a natural ceiling for the capabilities of these models.

[01:02:49.520]
Ned: No, never. More money, please. Local Stack snags 25 million for public cloud emulation. One of the great things about the public cloud is that it's pay as you go. One of the worst things about public cloud is it's pay as you go. If you've ever been tinkering around with AWS and accidentally left the NAT gateway running, you know what I'm talking about. You've got to be mindful of your resource usage when you're building for the cloud and make sure you clean up after yourself. There's also possible delays in getting things provisioned, especially for resource types that require approval for you to deploy. Local Stack aims to fix these problems for developers by bringing an emulated version of the cloud to their local workstation. Need an S3 bucket or an EKS cluster for testing your application? You can run and test it all locally with Local Stack without worrying about the hourly charge or the time it takes to deploy all those EC2 nodes. Local Stack currently supports a few hundred services from AWS, and their Azure and Snowflake models are in preview. Based on their success so far, they've managed to attract a $25 million Series A round led by notable capital.

[01:04:09.510]
Ned: While I was initially skeptical of what local stack is doing, it seems like they have a viable product, and it's providing useful benefits to developers. I fully expect them to get acquired by IBM in 2025.

[01:04:24.150]
Chris: Fbi announced this investigation into China-based cyberhacking against US telecommunications networks. You've heard of telecommunications, right? It's that thing that makes all the phones go, made by all those companies that everyone hates. Apparently, China hates them, too. A joint operation between the FBI and CISA, which stands for the Cyber Security and Infrastructure Security Agency, announced some fairly disturbing, if not terribly surprising findings. In short, hackers and hacking groups linked to Beijing are being accused of specifically targeting US telecoms in an effort to attack many individuals with close ties to, quote, government and political spheres. Considering the calendar, this is no surprise. This is just one of the many major hacking attacks that have been linked to China over the recent past, as the report also touches on... Well, for some reason, they're calling it Flax Typhoon. What?

[01:05:31.620]
Ned: Terrible.

[01:05:32.430]
Chris: Yeah, it's not great. What it is, is an attack platform that utilizes over 200,000 consumer devices, including cameras and routers across the United States. The number of WiFi-enabled pressure cookers that were found to be part of that botnet was not disclosed. But here at Chaos Lever, we strongly believe that it is in fact, not zero. Now, the idea that these attackers are going after telecom networks, specifically, is not surprising. It It has long been a fear that all of our infrastructure is poorly protected from a cybersecurity perspective, and the very nature of telecom networks makes them terminally online and therefore extremely juicy targets. China's government's response to the report was typical, basically saying, No, we did it. The report does not state any successes associated with the attack, which is not terribly surprising. But it does remind us that everyone from pressure cookerer owners to AT&T themselves has always got to be on guard, or I guess on guard, because we're not pronouncing any words right anymore.

[01:06:44.340]
Ned: Flax typhoon just sounds to me like a laxative or like a brand of cereal that you would buy, like Super Coal and blow. Dapper graduates the CNCF. Dapper was launched as an open-source project by Microsoft in in 2019. The goal of Dapper, spelled D-A-P-R because we hate E, is to create a distributed application run time for cloud-native applications, like those that run in Kubernetes or on functions as a service. These applications take a modular approach and are built from microservices that all need to communicate with each other. Dapper provides APIs for delivering that communication reliably and securely in a way that is vendor agnostic. You first define the what, like using a message broker, and then the how, using integrations from solutions like AWS SQS, Azure Service Bus, or Rabbit MQ. By decoupling the code from the implementation, your application is portable and able to adopt new services without rewriting or refactoring. Dapper joined the CNCF as an incubated project in November 2021, and in the last three years, it has hit the critical milestones of being stable, actively maintained, and achieving adoption across the industry. And so, as part of the PR maelstrom leading up to Cubecon North America, Dapper has officially reached graduated status.

[01:08:14.250]
Ned: No word on what it plans to wear to the ceremony, but one has to imagine it will be suitably fashionable.

[01:08:21.160]
Chris: Because it's-I got it.

[01:08:24.700]
Ned: All right, that's it. We're done now. Go away. Bye.