WEBVTT Kind: captions Language: en-US 00:00:00.000 --> 00:00:07.600 Hello and welcome to Distributed Systems and Blockchain in the News. My name is Thomas Bocek 00:00:07.600 --> 00:00:21.200 and today, this week, big news. The board of directors of OpenAI fired the CEO Sam Altman. 00:00:21.200 --> 00:00:33.760 Sam Altman, he co-founded OpenAI in 2015 and he became CEO in 2019 until basically yesterday. 00:00:33.760 --> 00:00:39.920 So the question, why is this related to this lecture, Distributed Systems and Blockchain? 00:00:39.920 --> 00:00:50.480 Well, first in 2019, Altman, he co-established Tools for Humanity, a venture that is developing 00:00:50.480 --> 00:00:59.680 a unique global biometric system based on iris recognition known as Worldcoin. And the system 00:00:59.680 --> 00:01:06.480 is designed to offer a dependable method for authenticating real human users 00:01:06.590 --> 00:01:12.950 on the internet with the objective to combating the challenges posed by AI 00:01:12.950 --> 00:01:18.390 generated fake identities and bots. Worldcoin's approach involves a 00:01:18.390 --> 00:01:23.270 cryptocurrency distribution model similar to universal basic income. 00:01:23.270 --> 00:01:30.230 However, keep in mind that Sam Bankman-Fried, the former CEO of FTX, promoted 00:01:30.230 --> 00:01:36.390 effective altruism while spending more than 300 million on properties, for 00:01:36.390 --> 00:01:43.830 example, more than 16 million dollars for a vacation home for his parents. 00:01:43.830 --> 00:01:49.950 But I have to say I'm not an expert in effective altruism. But more importantly 00:01:49.950 --> 00:01:58.030 it's disrupting IT at an incredible scale. It also affected my workflow and I 00:01:58.030 --> 00:02:04.310 can give you a couple of examples. So, for example, this video has subtitles which 00:02:04.310 --> 00:02:10.510 are created by Whisper, a large language model for creating subtitles. It takes a 00:02:10.510 --> 00:02:15.230 lot of CPU, so on my 12 core... 00:02:15.230 --> 00:02:21.020 machine it uses five minutes for a 10 minute video but it's superior to YouTube 00:02:21.020 --> 00:02:27.980 subtitle creator and I need to correct far less and this results in less time 00:02:27.980 --> 00:02:35.340 spending with creating subtitles. Another example is that one student group 00:02:35.340 --> 00:02:41.420 decided to use for their challenge task ICP as their blockchain in the challenge 00:02:41.420 --> 00:02:47.700 task and since I'm not an expert in ICP I created with ChatGPT I can show you 00:02:47.700 --> 00:02:55.940 here, based on the official documentation here, I create an expert system where one 00:02:55.940 --> 00:03:05.340 can ask questions about ICP and also this news segment ChatGPT helps me to 00:03:05.340 --> 00:03:11.620 summarize articles and those tasks are supported by large language models and 00:03:11.620 --> 00:03:17.860 these tasks can be finished quicker and one can focus on more important things. 00:03:17.860 --> 00:03:21.620 Large language models are not a silver bullet. For example, the summary of an article is often too large 00:03:21.620 --> 00:03:29.610 and the condensed version lacks the most important points 00:03:29.610 --> 00:03:34.090 I want to mention, but it reduces time I spend writing. 00:03:34.090 --> 00:03:40.370 So what happened with OpenAI and Sam Altman? 00:03:40.370 --> 00:03:43.610 Well, the board decided to fire Sam Altman 00:03:43.610 --> 00:03:45.690 and others resigned. 00:03:45.690 --> 00:03:48.250 For example, Greg Brockman, 00:03:48.250 --> 00:03:51.730 he tweeted also about it 00:03:51.730 --> 00:03:56.730 and other three senior researchers at OpenAI, 00:03:56.730 --> 00:03:59.450 they also quit. 00:03:59.450 --> 00:04:02.250 There are rumors about the reasons, 00:04:02.250 --> 00:04:04.490 but nothing is confirmed yet. 00:04:04.490 --> 00:04:07.970 Tweet of Greg Brockman ends with, 00:04:07.970 --> 00:04:11.930 "We will fine. Greater things coming soon." 00:04:11.930 --> 00:04:17.290 And one thing to mention is that OpenAI 00:04:17.290 --> 00:04:20.010 is still controversial. 00:04:20.010 --> 00:04:25.010 For example, the Dutch government in this article here. 00:04:28.010 --> 00:04:33.600 The Dutch government wants to ban officials from using AI software due to privacy and 00:04:33.600 --> 00:04:35.560 copyright concerns. 00:04:35.560 --> 00:04:41.400 I have to say from a productivity point of view, I would suggest the opposite. 00:04:41.400 --> 00:04:43.520 Make it mandatory. 00:04:43.520 --> 00:04:48.340 Use it. 00:04:48.340 --> 00:05:03.440 The next article discusses how [North] Korea has turned to various illicit means to earn 00:05:03.440 --> 00:05:08.640 foreign currency due to the international sanctions. 00:05:08.640 --> 00:05:18.140 Notably in 2022, North Korean hackers, they executed a massive cryptocurrency heist, stealing 00:05:18.140 --> 00:05:24.720 about 600 millions of the game Axie Infinity. 00:05:24.720 --> 00:05:31.300 And this was achieved through a fake job offer to a senior software engineer at the company, 00:05:31.300 --> 00:05:32.940 which allows the hackers to 00:05:32.950 --> 00:05:36.130 access private keys to the company's cryptocurrency account 00:05:36.130 --> 00:05:41.130 and then stealing more than 600 million in cryptocurrencies. 00:05:41.130 --> 00:05:45.550 Additionally, North Korea's hackers, in a significant shift, 00:05:45.550 --> 00:05:50.070 have focused on robbing cryptocurrency exchanges. 00:05:50.070 --> 00:05:53.630 Also in 2022, they were responsible 00:05:53.630 --> 00:05:57.190 for half of the cryptocurrency thefts, 00:05:57.190 --> 00:06:01.790 amassing around 1.6 billion US dollars. 00:06:01.790 --> 00:06:05.750 And this sum even surpassed North Korea's earning 00:06:05.750 --> 00:06:10.750 from all other experts and was significantly more 00:06:10.750 --> 00:06:14.110 than global earnings from ransomware. 00:06:14.110 --> 00:06:17.510 And the revenues from these cyber crimes 00:06:17.510 --> 00:06:20.870 have been used to fund missile tests 00:06:20.870 --> 00:06:25.870 and potentially assist Russia in the Ukraine conflict. 00:06:25.870 --> 00:06:28.750 And the article highlights the challenges 00:06:28.750 --> 00:06:31.310 in combat these activities. 00:06:31.310 --> 00:06:32.750 Very interesting article. 00:06:32.750 --> 00:06:39.710 The next article is from ArsTechnica. 00:06:39.710 --> 00:06:50.740 It's the following article here and it's about issues in software development. 00:06:50.740 --> 00:06:59.380 Despite ongoing warnings and guidance, many developers continue to embed sensitive 00:06:59.380 --> 00:07:06.820 credentials like cryptographic keys, security tokens, passwords directly into their publicly 00:07:06.820 --> 00:07:15.460 accessible source code and this practice has led to serious security breaches as for example 2015 00:07:15.460 --> 00:07:24.180 an incident at Uber where hackers accessed an internal database using a security key from a 00:07:24.180 --> 00:07:33.780 code shared publicly on github compromising data of 50'000 drivers and researchers from 00:07:33.780 --> 00:07:42.260 GitGuardian recently found almost 4'000 unique secrets in 450'000 projects submitted to 00:07:42.260 --> 00:07:45.860 PyPI, the python programming language's official 00:07:46.130 --> 00:07:55.250 code repository and these secrets exposed in various file types including mostly .py files, 00:07:55.250 --> 00:08:02.530 readmes, provided access to resources like Microsoft Active Directory, OAuth servers, 00:08:02.530 --> 00:08:11.570 SSH servers and third-party services. And the report indicated that the number of such exposures 00:08:11.570 --> 00:08:21.970 has increased over the past years. Despite the availability of safer practices and tools like 00:08:21.970 --> 00:08:30.690 .env files or AWS secret manager or secret manager from HashiCorp or code scanners, the issues 00:08:30.690 --> 00:08:37.970 persist even among large companies with strong security teams. The problem isn't confined to 00:08:37.970 --> 00:08:45.170 confined to Python or PyPI. Other open source repositories have historically faced similar issues and the article 00:08:45.170 --> 00:08:52.610 emphasizes that such exposure often accidentally poses significant risks enabling 00:08:52.720 --> 00:08:57.720 unauthorized access and potential exploitation. 00:08:57.720 --> 00:09:03.000 And another article, it's this article here. 00:09:03.000 --> 00:09:08.200 They analyzed stack overflow data. 00:09:08.200 --> 00:09:11.280 So this revealed also significant leaks 00:09:11.280 --> 00:09:15.120 of sensitive information primarily 00:09:15.120 --> 00:09:19.360 from users pasting log outputs 00:09:19.360 --> 00:09:23.960 or configuration files of various tools into their posts. 00:09:23.960 --> 00:09:28.960 And the researcher, after analyzing 00:09:28.960 --> 00:09:32.960 over 100 gigabyte of XML files 00:09:32.960 --> 00:09:36.040 from the stack exchange archive, 00:09:36.040 --> 00:09:38.280 discovered various leaks, 00:09:38.280 --> 00:09:43.280 including six valid GitHub user tokens 00:09:43.280 --> 00:09:45.480 out of 74 he examined. 00:09:45.480 --> 00:09:46.840 So he didn't examine all of it, 00:09:46.840 --> 00:09:50.720 but just out of those 74, six were actually valid. 00:09:50.720 --> 00:09:54.240 And most leaks originated not directly 00:09:54.240 --> 00:09:59.240 from code snippets brought from log outputs from tools. 00:09:59.310 --> 00:10:08.850 such as curl commands or packaged JSON files containing private tokens and this 00:10:08.850 --> 00:10:14.890 highlights the need for great caution in sharing information on public platforms 00:10:14.890 --> 00:10:22.830 like StackOverflow. So please be careful with keys and passwords. A good simple 00:10:22.830 --> 00:10:30.430 practice is to use .env files that are excluded from git and these files have 00:10:30.430 --> 00:10:41.270 to be shared out of band meaning not via git. The last article is about the 00:10:41.270 --> 00:10:48.910 messenger Signal. It's the following article here and this details the 00:10:48.910 --> 00:10:56.470 financial challenges of operating signal. Signal is a messenger client and this 00:10:56.470 --> 00:11:03.990 is a non-profit messaging app which is known for its robust encryption. 00:11:03.990 --> 00:11:07.790 Unlike profit-driven tech companies, 00:11:07.790 --> 00:11:14.060 Signal is funded mainly through donations facing substantial operating 00:11:14.060 --> 00:11:22.540 costs while prioritizing user privacy over monetization. Key expenses include infrastructure 00:11:22.540 --> 00:11:31.020 such as storage servers and bandwidth totaling approximately 14 million US dollars per year. 00:11:31.820 --> 00:11:39.580 Here down we have the cost and interestingly server cost is around 3 million US dollars 00:11:39.580 --> 00:11:46.540 per year. The bandwidth is also roughly 3 million US dollars per year. So the registration 00:11:46.540 --> 00:11:55.500 fee is 6 million which is roughly the value of the servers and the bandwidth combined. 00:11:57.100 --> 00:12:05.900 Also quite interesting in this article where they say they had 600 servers for private contact 00:12:05.900 --> 00:12:12.300 discovery and over time they reduced this to 10 servers so they downscaled and 00:12:12.490 --> 00:12:18.210 found optimizations to reduce this to only 10 servers. 00:12:18.210 --> 00:12:22.090 And additionally, the cost of human resources, 00:12:22.090 --> 00:12:26.210 crucial for maintaining and developing 00:12:26.210 --> 00:12:32.690 the app, is around $19 million US dollars per year. 00:12:32.690 --> 00:12:35.170 And despite these financial challenges, 00:12:35.170 --> 00:12:38.090 Signal remains committed to privacy, 00:12:38.090 --> 00:12:40.290 investing in advanced security features, 00:12:40.290 --> 00:12:42.910 and rejecting the surveillance-based business 00:12:42.910 --> 00:12:47.970 model, which we've seen in other tech industries. 00:12:47.970 --> 00:12:51.650 The author emphasizes the social importance 00:12:51.650 --> 00:12:56.410 of supporting privacy-focused alternative like Signal 00:12:56.410 --> 00:13:01.690 in a landscape of tech companies that dominate 00:13:01.690 --> 00:13:04.330 with data monetization. 00:13:04.330 --> 00:13:05.730 Very nice article. 00:13:05.730 --> 00:13:06.730 Take a look at it. 00:13:06.730 --> 00:13:11.730 This is a good insight into the financial situation of Signal. 00:13:11.730 --> 00:13:16.330 That's it from my side. 00:13:16.330 --> 00:13:18.330 See you next week.