1 00:00:00,000 --> 00:00:00,960  Welcome to Crash News.  2 00:00:00,960 --> 00:00:03,400  Today is May 20, 2025.  3 00:00:03,400 --> 00:00:06,120  Imagine having, say, a coding assistant  4 00:00:06,120 --> 00:00:08,280  that actually understands what you want  5 00:00:08,280 --> 00:00:10,400  and, well, just writes the code.  6 00:00:10,400 --> 00:00:14,520  Or a health system that spots the early whispers  7 00:00:14,520 --> 00:00:16,960  of a pandemic before it really explodes.  8 00:00:16,960 --> 00:00:18,520  Picture traffic flowing so smoothly  9 00:00:18,520 --> 00:00:20,520  you almost forget what a traffic jam feels like.  10 00:00:20,520 --> 00:00:22,600  It's a little bit like science fiction, right?  11 00:00:22,600 --> 00:00:25,000  - Maybe a few years ago, but not anymore.  12 00:00:25,000 --> 00:00:26,100  - Exactly.  13 00:00:26,100 --> 00:00:28,040  These things are actually starting to happen now  14 00:00:28,040 --> 00:00:31,280  thanks to some major breakthroughs in artificial intelligence.  15 00:00:31,280 --> 00:00:33,700  So today we're doing a deep dive into these advancements.  16 00:00:33,700 --> 00:00:35,600  We want to understand not just what's happening,  17 00:00:35,600 --> 00:00:38,240  but why it really matters to you.  18 00:00:38,240 --> 00:00:41,760  Think of this as your daily dose of AI education.  19 00:00:41,760 --> 00:00:44,920  We sift through all the noise, pull out the key insights,  20 00:00:44,920 --> 00:00:46,160  and try to make it engaging.  21 00:00:46,160 --> 00:00:47,140  - Right.  22 00:00:47,140 --> 00:00:49,160  We've gathered a whole stack of sources today,  23 00:00:49,160 --> 00:00:52,600  everything from healthcare innovation, transportation stuff,  24 00:00:52,600 --> 00:00:54,400  right through to the cutting edge of AI  25 00:00:54,400 --> 00:00:55,880  and software development.  26 00:00:55,880 --> 00:00:59,680  And AI vision tech too, that's growing incredibly fast.  27 00:00:59,680 --> 00:01:02,360  - Yeah, our mission today is really to cut through the hype,  28 00:01:02,360 --> 00:01:05,120  give you a clear picture of the real world impact.  29 00:01:05,120 --> 00:01:08,720  Okay, before we properly get started,  30 00:01:08,720 --> 00:01:11,160  a quick brain tickler for you.  31 00:01:11,160 --> 00:01:14,120  What's the name of OpenAI's newest AI tool,  32 00:01:14,120 --> 00:01:15,440  the one designed specifically  33 00:01:15,440 --> 00:01:16,800  to help with software engineering?  34 00:01:16,800 --> 00:01:18,440  - Good one, keep that in mind.  35 00:01:18,440 --> 00:01:20,500  - We'll reveal the answer later on.  36 00:01:20,500 --> 00:01:24,400  Now look, we know AI can sometimes sound, well,  37 00:01:24,400 --> 00:01:27,840  a bit technical like a foreign language almost.  38 00:01:27,840 --> 00:01:28,920  So whenever we hit a term  39 00:01:28,920 --> 00:01:31,000  that might need a little unpacking,  40 00:01:31,000 --> 00:01:33,400  we'll pause and explain it,  41 00:01:33,400 --> 00:01:34,840  make sure we're all on the same page.  42 00:01:34,840 --> 00:01:37,660  And if you find this kind of in-depth look valuable  43 00:01:37,660 --> 00:01:40,180  and you wanna help us keep it free and accessible,  44 00:01:40,180 --> 00:01:42,980  well, please take a moment to rate, like, and subscribe  45 00:01:42,980 --> 00:01:44,260  wherever you happen to be listening.  46 00:01:44,260 --> 00:01:45,480  - Yeah, that really helps.  47 00:01:45,480 --> 00:01:48,040  And we also have a growing Discord community,  48 00:01:48,040 --> 00:01:50,300  great place for deeper discussion on these topics.  49 00:01:50,300 --> 00:01:51,140  - Absolutely.  50 00:01:51,140 --> 00:01:52,680  And for those who wanna support us directly,  51 00:01:52,680 --> 00:01:56,740  maybe get some exclusive content, check out our Patreon.  52 00:01:56,740 --> 00:01:58,040  Think of this not just as news,  53 00:01:58,040 --> 00:02:00,460  but really as your ongoing education  54 00:02:00,460 --> 00:02:03,840  in this, well, rapidly changing world of AI.  55 00:02:03,840 --> 00:02:05,600  - Okay, let's get into it then.  56 00:02:05,600 --> 00:02:07,840  - Right, so the COVID-19 pandemic,  57 00:02:07,840 --> 00:02:11,540  it really, it's shown a spotlight on a critical need,  58 00:02:11,540 --> 00:02:12,380  didn't it?  59 00:02:12,380 --> 00:02:14,680  Better tools to see these health crises coming.  60 00:02:14,680 --> 00:02:15,500  - Absolutely.  61 00:02:15,500 --> 00:02:18,000  And it turns out AI is now being deployed  62 00:02:18,000 --> 00:02:19,040  right on that front line.  63 00:02:19,040 --> 00:02:21,760  It's analyzing just enormous amounts of information.  64 00:02:21,760 --> 00:02:22,960  Like, what kind of info?  65 00:02:22,960 --> 00:02:24,240  Well, everything.  66 00:02:24,240 --> 00:02:26,200  Global News Rep--  67 00:02:26,200 --> 00:02:32,230  reports, social media trends, airline passenger data, even traditional epidemiological figures,  68 00:02:32,230 --> 00:02:33,230  all pulled together.  69 00:02:33,230 --> 00:02:34,230  Wow.  70 00:02:34,230 --> 00:02:35,230  And the gold is what?  71 00:02:35,230 --> 00:02:36,230  Early detection.  72 00:02:36,230 --> 00:02:37,230  Exactly.  73 00:02:37,230 --> 00:02:38,710  The sheer volume of data.  74 00:02:38,710 --> 00:02:44,350  It's something humans just can't process effectively in real time, but AI can sift through it looking  75 00:02:44,350 --> 00:02:46,590  for those subtle patterns, those anomalies.  76 00:02:46,590 --> 00:02:49,670  That might signal an outbreak way before we'd normally spot it.  77 00:02:49,670 --> 00:02:50,670  Precisely.  78 00:02:50,670 --> 00:02:52,390  Weeks, maybe even months earlier.  79 00:02:52,390 --> 00:02:56,010  The idea is to give public health officials that crucial head start and also to model  80 00:02:56,010 --> 00:02:59,270  how a disease might spread, better decisions, faster responses.  81 00:02:59,270 --> 00:03:00,270  Makes sense.  82 00:03:00,270 --> 00:03:02,150  And it's not just about global health crises, is it?  83 00:03:02,150 --> 00:03:04,750  AI is also starting to reshape our daily journeys.  84 00:03:04,750 --> 00:03:05,750  Transportation.  85 00:03:05,750 --> 00:03:06,750  Yeah.  86 00:03:06,750 --> 00:03:08,950  There's a kind of quiet revolution happening there.  87 00:03:08,950 --> 00:03:10,490  Think about it.  88 00:03:10,490 --> 00:03:12,390  Traffic jams becoming rare.  89 00:03:12,390 --> 00:03:14,350  Buses arriving exactly on schedule.  90 00:03:14,350 --> 00:03:15,350  That would be nice.  91 00:03:15,350 --> 00:03:16,350  Wouldn't it?  92 00:03:16,350 --> 00:03:19,110  AI is really good at optimizing these complex systems.  93 00:03:19,110 --> 00:03:25,430  So in transport, you might have AI algorithms, dynamically changing traffic light timings  94 00:03:25,430 --> 00:03:27,790  based on real time flow.  95 00:03:27,790 --> 00:03:31,790  Or maybe predicting when a specific bus needs maintenance before it breaks down and causes  96 00:03:31,790 --> 00:03:32,870  delays.  97 00:03:32,870 --> 00:03:37,070  It's all about using data to make the whole network more efficient, more predictable.  98 00:03:37,070 --> 00:03:38,070  Right.  99 00:03:38,070 --> 00:03:39,070  Using data smarter.  100 00:03:39,070 --> 00:03:44,670  Now, let's shift to a, well, a truly massive challenge.  101 00:03:44,670 --> 00:03:45,870  Ending world hunger.  102 00:03:45,870 --> 00:03:47,990  It's a goal humanity's grappled with forever.  103 00:03:47,990 --> 00:03:49,430  A huge one.  104 00:03:49,430 --> 00:03:53,950  But there's growing optimism, apparently, that AI could be a really powerful tool here.  105 00:03:53,950 --> 00:03:57,230  Helping us tackle it, as some sources put it, smarter and faster.  106 00:03:57,230 --> 00:03:58,230  Yeah.  107 00:03:58,230 --> 00:03:59,790  But it's important to see AI as an enabler here, right?  108 00:03:59,790 --> 00:04:05,390  It can do amazing things like analyzing agricultural data to optimize crop yields, predict pests,  109 00:04:05,390 --> 00:04:06,790  streamline food supply chains.  110 00:04:06,790 --> 00:04:08,950  Minimize waste, that kind of thing.  111 00:04:08,950 --> 00:04:09,950  Exactly.  112 00:04:09,950 --> 00:04:13,630  But it can't solve the underlying socioeconomic factors that cause hunger.  113 00:04:13,630 --> 00:04:14,630  It's not a magic bullet.  114 00:04:14,630 --> 00:04:15,630  No, of course not.  115 00:04:15,630 --> 00:04:20,450  But its ability to process and make sense of all the complex data involved in food production  116 00:04:20,450 --> 00:04:24,510  and distribution, that makes it pretty much indispensable in the global effort.  117 00:04:24,510 --> 00:04:25,510  Okay.  118 00:04:25,510 --> 00:04:28,590  Shifting gears again, let's talk about information itself.  119 00:04:28,590 --> 00:04:29,590  Journalism.  120 00:04:29,590 --> 00:04:32,330  AI is popping up more and more in newsrooms.  121 00:04:32,330 --> 00:04:33,330  It is.  122 00:04:33,330 --> 00:04:37,510  For tasks like transcribing interviews, which takes ages manually.  123 00:04:37,510 --> 00:04:39,270  Or translating articles.  124 00:04:39,270 --> 00:04:40,990  Even generating basic news reports.  125 00:04:40,990 --> 00:04:43,110  Like weather or financial summaries.  126 00:04:43,110 --> 00:04:44,110  Yeah.  127 00:04:44,110 --> 00:04:45,910  Like quality reports, things like that.  128 00:04:45,910 --> 00:04:48,630  Factual, data-driven stuff.  129 00:04:48,630 --> 00:04:50,270  But this obviously brings up some...  130 00:04:50,300 --> 00:04:52,340  - Well, pretty important ethical question.  131 00:04:52,340 --> 00:04:53,560  - Of what, accuracy?  132 00:04:53,560 --> 00:04:54,400  Bias?  133 00:04:54,400 --> 00:04:55,220  - Exactly.  134 00:04:55,220 --> 00:04:57,820  How do you ensure AI-generated content is accurate,  135 00:04:57,820 --> 00:05:00,140  transparent, and free from bias?  136 00:05:00,140 --> 00:05:03,940  Who's responsible if an AI makes a mistake in a news story?  137 00:05:03,940 --> 00:05:04,780  - Tricky questions.  138 00:05:04,780 --> 00:05:05,620  - Very.  139 00:05:05,620 --> 00:05:07,420  It's about balancing the efficiency gains  140 00:05:07,420 --> 00:05:09,460  with the core principles of journalism.  141 00:05:09,460 --> 00:05:12,260  It frees up journalists for deeper work, which is good,  142 00:05:12,260 --> 00:05:14,140  but the oversight needs to be there.  143 00:05:14,140 --> 00:05:14,980  - Right.  144 00:05:14,980 --> 00:05:16,220  It feels like AI's been around,  145 00:05:16,220 --> 00:05:19,540  but the pace of change right now seems different.  146 00:05:19,540 --> 00:05:21,820  We hear terms like exponential growth  147 00:05:21,820 --> 00:05:23,500  after those AI winters.  148 00:05:23,500 --> 00:05:25,020  - Yeah, those winters were periods  149 00:05:25,020 --> 00:05:28,720  where progress kind of stalled, funding dried up.  150 00:05:28,720 --> 00:05:30,020  Overcoming those was key.  151 00:05:30,020 --> 00:05:33,300  It was driven by big leaps in computing power  152 00:05:33,300 --> 00:05:35,540  and crucially, the availability  153 00:05:35,540 --> 00:05:37,180  of massive data sets for training.  154 00:05:37,180 --> 00:05:40,260  - So this exponential growth, it's not just faster,  155 00:05:40,260 --> 00:05:42,060  it's fundamentally changing things.  156 00:05:42,060 --> 00:05:42,900  - I think so.  157 00:05:42,900 --> 00:05:46,380  It's changing how we solve problems across so many fields.  158 00:05:46,380 --> 00:05:47,820  Medical science, for example,  159 00:05:47,820 --> 00:05:51,460  AI speeding up drug discovery, improving diagnostics,  160 00:05:51,460 --> 00:05:53,580  even in literature, analyzing texts,  161 00:05:53,580 --> 00:05:55,300  exploring new creative forms.  162 00:05:55,300 --> 00:05:56,340  - Fascinating.  163 00:05:56,340 --> 00:05:57,500  Okay, let's move to an area  164 00:05:57,500 --> 00:05:58,960  maybe not everyone thinks about daily,  165 00:05:58,960 --> 00:06:00,580  but it's vital for business.  166 00:06:00,580 --> 00:06:01,740  Treasury operations.  167 00:06:01,740 --> 00:06:03,300  - Ah, yes, managing the money.  168 00:06:03,300 --> 00:06:04,120  - Basically, yeah.  169 00:06:04,120 --> 00:06:05,340  Cashflow, making sure companies  170 00:06:05,340 --> 00:06:07,460  have enough liquid cash on hand.  171 00:06:07,460 --> 00:06:09,020  Experts are saying AI and automation  172 00:06:09,020 --> 00:06:10,380  are really shaking things up here.  173 00:06:10,380 --> 00:06:11,220  - They are.  174 00:06:11,220 --> 00:06:13,500  Traditionally, treasury involves a lot of manual work,  175 00:06:13,500 --> 00:06:15,380  spreadsheets, forecasts.  176 00:06:15,380 --> 00:06:18,860  It's time consuming and frankly, prone to errors.  177 00:06:18,860 --> 00:06:21,640  - So AI steps in to automate the routine stuff?  178 00:06:21,640 --> 00:06:22,760  - Exactly.  179 00:06:22,760 --> 00:06:24,000  Automate routine tasks,  180 00:06:24,000 --> 00:06:27,540  but also provide much more sophisticated data analytics.  181 00:06:27,540 --> 00:06:30,620  The goal is better speed, better accuracy,  182 00:06:30,620 --> 00:06:31,860  better liquidity management.  183 00:06:31,860 --> 00:06:32,940  - How specifically?  184 00:06:32,940 --> 00:06:35,460  Cashflow forecasting is one area mentioned.  185 00:06:35,460 --> 00:06:36,300  - Right.  186 00:06:36,300 --> 00:06:38,060  Instead of relying on historical trends  187 00:06:38,060 --> 00:06:40,780  and let's be honest, sometimes educated guesses,  188 00:06:40,780 --> 00:06:43,820  AI can analyze vast amounts of past financial data  189 00:06:43,820 --> 00:06:46,940  to predict future cashflow much more accurately.  190 00:06:46,940 --> 00:06:49,120  - Going from reactive to predictive.  191 00:06:49,120 --> 00:06:50,260  - That's the key phrase.  192 00:06:50,260 --> 00:06:52,500  It lets treasury teams anticipate shortfalls  193 00:06:52,500 --> 00:06:54,140  or surpluses much earlier.  194 00:06:54,140 --> 00:06:56,420  Better planning, more strategic decisions.  195 00:06:56,420 --> 00:06:57,780  - And it's not just predicting the future,  196 00:06:57,780 --> 00:06:59,540  it's about the now too, right?  197 00:06:59,540 --> 00:07:00,820  Real-time data access.  198 00:07:00,820 --> 00:07:01,860  - Absolutely.  199 00:07:01,860 --> 00:07:05,920  AI allows for an up-to-the-minute view of cash positions.  200 00:07:05,920 --> 00:07:08,820  In today's fast markets, waiting for end-of-day reports  201 00:07:08,820 --> 00:07:09,820  just doesn't cut it anymore.  202 00:07:09,820 --> 00:07:11,480  You need that real-time visibility  203 00:07:11,480 --> 00:07:13,880  to make quick, informed decisions.  204 00:07:13,880 --> 00:07:14,720  Be agile.  205 00:07:14,720 --> 00:07:15,560  - Makes sense.  206 00:07:15,560 --> 00:07:17,590  and risk management, fraud detection.  207 00:07:17,590 --> 00:07:18,650  - Another big one.  208 00:07:18,650 --> 00:07:20,950  Financial fraud is getting more sophisticated.  209 00:07:20,950 --> 00:07:23,710  AI can analyze massive volumes of transactions,  210 00:07:23,710 --> 00:07:27,390  spot anomalies, flag suspicious activity in real time.  211 00:07:27,390 --> 00:07:29,170  It's a crucial layer of security.  212 00:07:29,170 --> 00:07:31,390  - So the takeaway for treasury seems to be,  213 00:07:31,390 --> 00:07:34,590  embrace AI and automation for better efficiency,  214 00:07:34,590 --> 00:07:36,710  less risk and smarter decisions.  215 00:07:36,710 --> 00:07:37,550  - Pretty much.  216 00:07:37,550 --> 00:07:40,710  It's still relatively early days for widespread adoption,  217 00:07:40,710 --> 00:07:43,110  but the potential transformation is huge.  218 00:07:43,110 --> 00:07:44,510  Companies that get on board now  219 00:07:44,510 --> 00:07:46,590  will likely be much better positioned.  220 00:07:46,590 --> 00:07:47,790  - Okay, let's widen the lens again.  221 00:07:47,790 --> 00:07:49,090  Computer vision.  222 00:07:49,090 --> 00:07:53,130  AI that lets computers see and interpret images.  223 00:07:53,130 --> 00:07:56,430  The market growth here is just explosive.  224 00:07:56,430 --> 00:07:58,110  - The numbers are staggering, aren't they?  225 00:07:58,110 --> 00:08:01,270  Valued at over $24 billion in 2023.  226 00:08:01,270 --> 00:08:05,350  - Projected to hit nearly $174 billion by 2032.  227 00:08:05,350 --> 00:08:08,790  That's a compound annual growth rate of almost 25%.  228 00:08:08,790 --> 00:08:11,390  - Yeah, it really shows the power of this technology.  229 00:08:11,390 --> 00:08:14,430  It's being driven by the demand for automation everywhere,  230 00:08:14,430 --> 00:08:16,550  huge improvements in image recognition,  231 00:08:16,550 --> 00:08:19,190  thanks to AI, deep learning advancements.  232 00:08:19,190 --> 00:08:20,910  - And it's impacting key industries, right?  233 00:08:20,910 --> 00:08:22,350  Healthcare, automotive.  234 00:08:22,350 --> 00:08:23,190  - Definitely.  235 00:08:23,190 --> 00:08:25,710  Healthcare for things like detecting diseases earlier  236 00:08:25,710 --> 00:08:28,110  from scans, more precise diagnoses.  237 00:08:28,110 --> 00:08:30,750  Automotive has massive driver assistance systems,  238 00:08:30,750 --> 00:08:31,730  autonomous vehicles.  239 00:08:31,730 --> 00:08:33,670  They rely entirely on computer vision.  240 00:08:33,670 --> 00:08:35,710  - We're also seeing it in smart retail, surveillance.  241 00:08:35,710 --> 00:08:36,550  - Right.  242 00:08:36,550 --> 00:08:38,050  Retail uses it for inventory,  243 00:08:38,050 --> 00:08:39,710  understanding customer behavior,  244 00:08:39,710 --> 00:08:41,430  surveillance for security.  245 00:08:41,430 --> 00:08:43,150  The applications are incredibly broad.  246 00:08:43,150 --> 00:08:45,110  - And the big tech players are all involved.  247 00:08:45,110 --> 00:08:47,870  Nvidia, Intel, Microsoft, Google.  248 00:08:47,870 --> 00:08:49,830  - Yeah, they're the ones with the R&D muscle,  249 00:08:49,830 --> 00:08:51,590  the infrastructure, the funding.  250 00:08:51,590 --> 00:08:54,110  Nvidia's GPUs, for instance, are critical.  251 00:08:54,110 --> 00:08:55,990  But it's also the whole ecosystem.  252 00:08:55,990 --> 00:08:58,750  Research labs, startups getting funding.  253 00:08:58,750 --> 00:09:00,630  It all fuels this growth.  254 00:09:00,630 --> 00:09:02,890  - Okay, now let's zero in on something specific  255 00:09:02,890 --> 00:09:05,070  that's generating a lot of buzz.  256 00:09:05,070 --> 00:09:07,030  Open AI's Codex.  257 00:09:07,030 --> 00:09:08,710  This sounds like it could really change  258 00:09:08,710 --> 00:09:10,150  how software gets made.  259 00:09:10,150 --> 00:09:12,070  - It's definitely positioned as a major step.  260 00:09:12,070 --> 00:09:15,030  The core idea is AI that understands instructions  261 00:09:15,030 --> 00:09:17,310  in plain English and then generates the code,  262 00:09:17,310 --> 00:09:19,470  making coding more intuitive, potentially.  263 00:09:19,470 --> 00:09:20,590  - More than just auto-complete.  264 00:09:20,590 --> 00:09:21,470  - Much more.  265 00:09:21,470 --> 00:09:23,390  It's aiming to bridge that gap  266 00:09:23,390 --> 00:09:25,370  between what a developer wants to do  267 00:09:25,370 --> 00:09:28,070  and the complex syntax needed to actually do it.  268 00:09:28,070 --> 00:09:30,670  - We saw notes from PixelCrayons highlighting the limits  269 00:09:30,670 --> 00:09:33,070  of traditional coding time, errors,  270 00:09:33,070 --> 00:09:35,910  and seeing AI tools like Codex as the solution.  271 00:09:35,910 --> 00:09:36,750  - Right.  272 00:09:36,750 --> 00:09:38,310  And they mention Emma Joseph,  273 00:09:38,310 --> 00:09:40,470  specializing in blockchain and AI,  274 00:09:40,600 --> 00:09:42,400  which points to how these advanced techs are  275 00:09:42,400 --> 00:09:44,640  merging to create new solutions.  276 00:09:44,640 --> 00:09:48,560  codecs streamlining coding could massively boost efficiency.  277 00:09:48,560 --> 00:09:51,300  There's also a specific project, codecs-cli,  278 00:09:51,300 --> 00:09:53,040  for developers who work in the terminal.  279 00:09:53,040 --> 00:09:54,720  Yeah, command line interface.  280 00:09:54,720 --> 00:09:57,720  It basically brings chat GPT level reasoning  281 00:09:57,720 --> 00:09:59,440  into the terminal workflow.  282 00:09:59,440 --> 00:10:01,760  It can run code, manage files, work  283 00:10:01,760 --> 00:10:05,760  with Git for version control, all within a secure sandbox.  284 00:10:05,760 --> 00:10:06,780  Sounds powerful.  285 00:10:06,780 --> 00:10:10,080  It can refactor code, generate database migrations,  286 00:10:10,080 --> 00:10:11,520  write unit tests.  287 00:10:11,520 --> 00:10:14,600  Even explain complex things like regular expressions,  288 00:10:14,600 --> 00:10:16,980  suggest improvements via poll requests,  289 00:10:16,980 --> 00:10:18,400  do basic security reviews.  290 00:10:18,400 --> 00:10:20,800  These are all tasks that eat up developer time.  291 00:10:20,800 --> 00:10:22,500  Automating the grunt work, potentially.  292 00:10:22,500 --> 00:10:24,200  That's the idea, freeing up developers  293 00:10:24,200 --> 00:10:25,560  for higher level thinking.  294 00:10:25,560 --> 00:10:27,360  And it works with standard tools like Git,  295 00:10:27,360 --> 00:10:29,980  which is important for fitting into existing workflows.  296 00:10:29,980 --> 00:10:32,760  The requirements seem fairly standard in Node.js, major OS  297 00:10:32,760 --> 00:10:33,600  support.  298 00:10:33,600 --> 00:10:36,320  And you can configure it, choose different AI models.  299 00:10:36,320 --> 00:10:37,540  Yeah, that's a plus.  300 00:10:37,540 --> 00:10:40,920  It supports models from OpenAI, Azure, Google, even  301 00:10:40,920 --> 00:10:44,160  open source options, gives developers flexibility.  302 00:10:44,160 --> 00:10:47,080  You can tweak how it asks for approval, handles errors,  303 00:10:47,080 --> 00:10:48,400  sends notifications.  304 00:10:48,400 --> 00:10:50,860  And OpenAI is encouraging others to contribute.  305 00:10:50,860 --> 00:10:51,740  Which is smart.  306 00:10:51,740 --> 00:10:54,480  Community involvement helps find bugs, add features,  307 00:10:54,480 --> 00:10:57,360  make it more robust, especially for an experimental tool  308 00:10:57,360 --> 00:10:58,300  like this.  309 00:10:58,300 --> 00:11:01,680  OK, looking at other news sources, just AI news,  310 00:11:01,680 --> 00:11:03,360  mentioned some big funding rounds.  311 00:11:03,360 --> 00:11:07,480  $43 million for Granola, an AI workspace.  312 00:11:07,480 --> 00:11:11,740  And $6 million for Plexus, for AI legal automation.  313 00:11:11,740 --> 00:11:14,060  Yeah, that shows the investment pouring into AI  314 00:11:14,060 --> 00:11:15,560  across different sectors.  315 00:11:15,560 --> 00:11:18,020  Collaboration tools, legal tech.  316 00:11:18,020 --> 00:11:20,580  AI is seen as a key driver pretty much everywhere.  317 00:11:20,580 --> 00:11:23,940  Just AI news also keeps an eye on the broader picture impact  318 00:11:23,940 --> 00:11:27,940  across cybersecurity, finance, but also the ethics,  319 00:11:27,940 --> 00:11:29,320  policy privacy stuff.  320 00:11:29,320 --> 00:11:30,160  Which is vital.  321 00:11:30,160 --> 00:11:31,420  It's not just about the tech.  322 00:11:31,420 --> 00:11:34,560  It's the whole societal context essential to track that.  323 00:11:34,560 --> 00:11:36,580  Over on YouTube, the AI Revolution channel  324 00:11:36,580 --> 00:11:38,440  covered codecs too, confirming it's  325 00:11:38,440 --> 00:11:41,520  a research preview for pro, team, and enterprise users.  326 00:11:41,520 --> 00:11:43,280  Right, integrated into chat GPT.  327 00:11:43,280 --> 00:11:45,800  Makes sense to roll it out that way first, gather feedback.  328 00:11:45,800 --> 00:11:49,040  They described it as a full stack dev in a secure cloud  329 00:11:49,040 --> 00:11:50,520  sandbox.  330 00:11:50,520 --> 00:11:54,320  No external internet access, but it connects to GitHub repos.  331 00:11:54,320 --> 00:11:56,340  Yeah, the autonomy is key.  332 00:11:56,340 --> 00:11:59,240  It can apparently handle tasks, end-to-end writing features,  333 00:11:59,240 --> 00:12:02,140  fixing bugs, running tests autonomously.  334 00:12:02,140 --> 00:12:05,200  The secure sandbox is crucial for protecting code.  335 00:12:05,200 --> 00:12:06,040  And you can watch.  336 00:12:06,040 --> 00:12:11,910  it work in real time? Logs, test results. Transparency, yeah. See what it's doing.  337 00:12:11,910 --> 00:12:16,950  It's powered by a model called Codex One, a version of O3 fine-tuned specifically for  338 00:12:16,950 --> 00:12:20,650  coding using reinforcement learning. Trained on real coding tasks.  339 00:12:20,650 --> 00:12:24,150  Which hopefully makes it generate more practical, effective code.  340 00:12:24,150 --> 00:12:29,670  They mentioned benchmarks too. Codex One apparently hitting 75% pass rate on some tasks. Better  341 00:12:29,670 --> 00:12:34,190  than the standard O3 high model. That's a measurable improvement. Shows progress  342 00:12:34,190 --> 00:12:37,950  in generating reliable code. And apparently it needs minimal setup, though you can give  343 00:12:37,950 --> 00:12:40,510  it hints with special files to navigate repos better.  344 00:12:40,510 --> 00:12:47,030  AI revolution also touched on anthropic, potentially updating Claude, maybe towards true agentic  345 00:12:47,030 --> 00:12:51,010  behavior. Ah, yes. Agents that can reason and act autonomously,  346 00:12:51,010 --> 00:12:56,030  formulate plans, use tools, iterate with less direct prompting. That's a big trend.  347 00:12:56,030 --> 00:13:01,150  And Google, of course, putting AI into search. Overviews are here. An AI mode coming.  348 00:13:01,150 --> 00:13:06,110  Great conversational search powered by Gemini. Changing how we find information online. It's  349 00:13:06,110 --> 00:13:09,790  a response to these powerful conversational AIs.  350 00:13:09,790 --> 00:13:15,310  Another YouTuber, David Andrzej, was even more bullish. Called Codex the most powerful  351 00:13:15,310 --> 00:13:19,270  AI coding agent ever created. A threat to existing tools.  352 00:13:19,270 --> 00:13:24,990  Strong words. But it highlights the potential disruption. And putting it right inside ChatGPT  353 00:13:24,990 --> 00:13:28,310  makes it super accessible. He stressed it's cloud-based. Asynchronous  354 00:13:28,310 --> 00:13:34,070  handles multiple tasks at once. Drafting pull requests, finding bugs, reviewing code, writing  355 00:13:34,070 --> 00:13:36,470  tests. Yeah, the cloud power lets it do heavy lifting  356 00:13:36,470 --> 00:13:39,870  concurrently. Covering so much of the development lifecycle is impressive.  357 00:13:39,870 --> 00:13:44,390  He also mentioned the slightly better benchmarks for Codex One and showed it working on a real  358 00:13:44,390 --> 00:13:47,950  production code base. Seeing it in action on real code is always  359 00:13:47,950 --> 00:13:51,670  more convincing than just benchmarks. And the remote operation thing using it from  360 00:13:51,670 --> 00:13:55,430  your phone because it runs in a virtual environment in OpenAI's cloud.  361 00:13:55,430 --> 00:13:59,550  That's a big flexibility win for developers. Work from anywhere. The isolated environment  362 00:13:59,550 --> 00:14:02,950  keeps things clean. India Today echoed much of this accessible  363 00:14:02,950 --> 00:14:09,690  via ChatGPT. Tasks take 1-30 minutes, real-time monitoring, isolated workspaces, ability to  364 00:14:09,690 --> 00:14:13,790  modify files and run commands. Consistent reporting. The picture of how  365 00:14:13,790 --> 00:14:17,910  it operates is becoming clearer. It's not just code generation. It's interacting with  366 00:14:17,910 --> 00:14:21,390  the dev environment. ZDNet framed it as enabling a real coding  367 00:14:21,390 --> 00:14:25,910  mindset change. Beyond autocomplete to AI writing whole segments  368 00:14:25,910 --> 00:14:29,310  analyzing systems. That's the potential shift. If AI handles  369 00:14:29,310 --> 00:14:30,670  more complexity reliably.  370 00:14:30,900 --> 00:14:34,100  Developers focus shifts to architecture, high level design.  371 00:14:34,100 --> 00:14:36,220  - But ZDNet also raised concerns.  372 00:14:36,220 --> 00:14:38,180  Impact on junior programmers.  373 00:14:38,180 --> 00:14:41,100  Losing learning opportunities if you rely too much on AI.  374 00:14:41,100 --> 00:14:42,060  - Very valid points.  375 00:14:42,060 --> 00:14:43,420  Productivity gains are great,  376 00:14:43,420 --> 00:14:44,880  but what about entry level roles?  377 00:14:44,880 --> 00:14:46,480  How do people learn fundamentals?  378 00:14:46,480 --> 00:14:48,000  The industry needs to think about that.  379 00:14:48,000 --> 00:14:49,620  Training will have to adapt.  380 00:14:49,620 --> 00:14:52,260  - Singularity Hub just summed it up neatly.  381 00:14:52,260 --> 00:14:54,460  Translates everyday language into computer code.  382 00:14:54,460 --> 00:14:55,660  - Gets to the core of it, yeah.  383 00:14:55,660 --> 00:14:57,420  Making programming more accessible.  384 00:14:57,420 --> 00:15:00,660  - Looking at Reddit discussions, a real mix of reactions.  385 00:15:01,500 --> 00:15:02,860  Excitement, definitely.  386 00:15:02,860 --> 00:15:05,460  But also frustration that it's a limited preview,  387 00:15:05,460 --> 00:15:07,740  not available for regular Plus users yet.  388 00:15:07,740 --> 00:15:08,720  - Totally understandable.  389 00:15:08,720 --> 00:15:11,320  High anticipation meets controlled rollout.  390 00:15:11,320 --> 00:15:12,940  Happens all the time with new tech.  391 00:15:12,940 --> 00:15:15,500  - There was also debate about the name OpenAI,  392 00:15:15,500 --> 00:15:17,140  given its commercial direction.  393 00:15:17,140 --> 00:15:19,060  And questions about limitations.  394 00:15:19,060 --> 00:15:22,900  Can it really run and test complex apps?  395 00:15:22,900 --> 00:15:25,540  Concerns about cloud reliance, GitHub integration.  396 00:15:25,540 --> 00:15:26,380  - All fair questions.  397 00:15:26,380 --> 00:15:28,640  The open versus commercial tension is ongoing.  398 00:15:28,640 --> 00:15:30,300  People want to know the practical limits,  399 00:15:30,300 --> 00:15:31,500  the dependencies.  400 00:15:31,500 --> 00:15:33,500  - But some Redditors who do have access  401 00:15:33,500 --> 00:15:34,960  are calling it amazing.  402 00:15:34,960 --> 00:15:36,820  Lots of calls for a VSCode plug-in too.  403 00:15:36,820 --> 00:15:39,300  - Early positive feedback is promising.  404 00:15:39,300 --> 00:15:41,700  And the VSCode request makes total sense.  405 00:15:41,700 --> 00:15:43,540  Integration is key for adoption.  406 00:15:43,540 --> 00:15:44,380  Of course, you also get  407 00:15:44,380 --> 00:15:46,780  the usual unsubstantiated claims online.  408 00:15:46,780 --> 00:15:47,620  - Right.  409 00:15:47,620 --> 00:15:50,820  Okay, last big chunk of news was a compilation.  410 00:15:50,820 --> 00:15:54,740  Some odd bits like a disappearing Alexa Plus news,  411 00:15:54,740 --> 00:15:57,740  mention of AI art by Chris Castanova,  412 00:15:57,740 --> 00:15:59,380  and Google's AlphaVolve.  413 00:15:59,380 --> 00:16:00,740  - AlphaVolve sounds fascinating.  414 00:16:00,740 --> 00:16:03,520  An AI that invent solutions, improve data centers,  415 00:16:03,520 --> 00:16:05,640  AI chips sped up Gemini training.  416 00:16:05,640 --> 00:16:07,940  That's potentially groundbreaking if it scales.  417 00:16:07,940 --> 00:16:09,540  - Then there was Microsoft's build event,  418 00:16:09,540 --> 00:16:12,900  open agentic web vision, co-pilot updates, new tools,  419 00:16:12,900 --> 00:16:14,580  but also those Microsoft layoffs  420 00:16:14,580 --> 00:16:17,300  hitting software engineers, leading to speculation.  421 00:16:17,300 --> 00:16:19,280  - Yeah, the timing raises questions, doesn't it?  422 00:16:19,280 --> 00:16:21,780  While not explicitly linked by Microsoft,  423 00:16:21,780 --> 00:16:23,000  people wonder about the connection  424 00:16:23,000 --> 00:16:26,000  between advancing AI coding tools and engineering rules.  425 00:16:26,000 --> 00:16:27,180  It's a complex issue.  426 00:16:27,180 --> 00:16:31,740  Also mentioned, multi-agent stuff and M365 co-pilot,  427 00:16:31,740 --> 00:16:36,100  quad code SDK, NVIDIA's big AI campus plans for Europe,  428 00:16:36,100 --> 00:16:39,180  the SAG-AFTRA complaint about the AI Darth Vader voice.  429 00:16:39,180 --> 00:16:40,540  - That voice issue is a big one  430 00:16:40,540 --> 00:16:41,980  for actors and voice artists,  431 00:16:41,980 --> 00:16:44,180  shows the IP and ethical challenges spreading.  432 00:16:44,180 --> 00:16:45,940  - The study showing GPT-4 winning debates  433 00:16:45,940 --> 00:16:47,500  more often with demographic data,  434 00:16:47,500 --> 00:16:49,900  funding for Tensorwave and some might AI,  435 00:16:49,900 --> 00:16:52,860  AI jobs at Meta and Reddit, prompt tutorials,  436 00:16:52,860 --> 00:16:55,200  Gemini on Nest Audio, loads of stuff.  437 00:16:55,200 --> 00:16:57,340  It really highlights--  438 00:16:57,340 --> 00:17:01,970  the sheer breadth of AI activity. Enterprise, creative, ethics, hardware, jobs. It's touching  439 00:17:01,970 --> 00:17:06,650  everything. Even that anecdote about O3 as a career coach. Unexpected applications.  440 00:17:06,650 --> 00:17:12,430  We also saw info on top open source large language models for 2025. Highlighting benefits  441 00:17:12,430 --> 00:17:19,090  like cost, customization, transparency, community innovation, data security. Models like Laminae  442 00:17:19,090 --> 00:17:23,450  3, Gemma 2, Command R, Plus Day. Yeah, open source is hugely important. It  443 00:17:23,450 --> 00:17:28,370  democratizes access. Companies aren't solely reliant on proprietary models. The benefits  444 00:17:28,370 --> 00:17:31,970  you listed are real advantages. The discussion covered innovations in training  445 00:17:31,970 --> 00:17:37,490  data, model architecture, tuning techniques, multimodal abilities, plus responsible deployment,  446 00:17:37,490 --> 00:17:43,050  ethics, safety, community support via Hugging Face and GitHub. Future trends like multilingualism,  447 00:17:43,050 --> 00:17:46,450  efficiency, hybrid solutions, even guidance on building your own.  448 00:17:46,450 --> 00:17:50,250  It's a whole ecosystem. Great to see that depth. Looking not just at the tech, but also  449 00:17:50,250 --> 00:17:54,930  the responsible use and community aspects. Empowering people to build and customize is  450 00:17:54,930 --> 00:17:56,990  key. And finally, that report about a potential  451 00:17:56,990 --> 00:18:02,210  10-year block on state level AI regulation enforcement in the US budget bill. With pushback  452 00:18:02,210 --> 00:18:06,410  from state AGs and groups like Mozilla, EPEC. That's a major potential development on the  453 00:18:06,410 --> 00:18:12,930  regulation front. The classic tension. Innovation versus consumer protection and safety. Limiting  454 00:18:12,930 --> 00:18:18,170  state oversight could have big consequences given how fast AI is moving. The outcome is  455 00:18:18,170 --> 00:18:21,510  uncertain, it sounds like. And just to note that UT Austin is proposing  456 00:18:21,510 --> 00:18:25,770  AI guidelines for teaching and learning. So education is grappling with this too.  457 00:18:25,770 --> 00:18:30,890  Absolutely. Universities, schools, everyone needs to figure out how to integrate AI responsibly  458 00:18:30,890 --> 00:18:36,110  and ethically. So wrapping this all up, what's the big picture?  459 00:18:36,110 --> 00:18:41,530  We've seen AI moving fast from theory to reality. Predicting pandemics, automating coding with  460 00:18:41,530 --> 00:18:47,530  tools like OpenAI's Codex, which yes, was the pop tis answer. The pace is just staggering.  461 00:18:47,530 --> 00:18:52,730  It really is. But with great power comes, well, crucial questions about ethics, regulation,  462 00:18:52,730 --> 00:18:57,530  the impact on jobs, on how we live. That debate over state versus federal regulation really  463 00:18:57,530 --> 00:19:00,830  highlights the core tensions. This deep dive, as always, just scratches  464 00:19:00,830 --> 00:19:04,570  the surface. We definitely encourage you to explore the sources further. What stands out  465 00:19:04,570 --> 00:19:07,210  most to you listening? What benefits or risks did we miss?  466 00:19:07,210 --> 00:19:12,330  Yeah, definitely keep exploring. And maybe a final thought to chew on, as AI keeps evolving,  467 00:19:12,330 --> 00:19:16,950  maybe even learning to invent its own solutions like Alpha Evolve, how does our fundamental  468 00:19:16,950 --> 00:19:21,030  relationship with technology change? And what new responsibilities does that place on all of us?