Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Artificial Intelligence CW 41/ 42

Transcript

Best of LinkedIn: Artificial Intelligence CW 41/ 42 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 1: This episode is provided by Thomas Allgaier and Frenes based on the most relevant LinkedIn posts about artificial intelligence in calendar weeks, forty one and forty two.
• 0:09
  Speaker 1: Frenes supports enterprises with market and competitive intelligence, decoding emerging technologies, customer insights, regulatory shifts and competitor strategies.
• 0:18
  Speaker 1: So product teams and strategy leaders don't just react, but shape the future of AI.
• 0:24
  Speaker 0: Welcome back.
• 0:26
  Speaker 0: So today we're really going to dig into the top artificial intelligence trends we've been seeing across professional insights from the last couple of weeks.
• 0:34
  Speaker 0: Right.
• 0:34
  Speaker 0: The mission really is to figure out where things are shifting, you know, moving past the hype, the pilots towards disciplined, measurable execution, real enterprise stuff.
• 0:43
  Speaker 1: Exactly.
• 0:44
  Speaker 1: Execution, that's the word.
• 0:45
  Speaker 1: Yeah.
• 0:45
  Speaker 1: The conversation feels like it's grown up a bit.
• 0:48
  Speaker 1: We've kind of clustered the insights for you today into three main areas.
• 0:51
  Speaker 1: Okay.
• 0:52
  Speaker 1: We'll kick off with the foundation infrastructure, the chip wars, up and how these new strategic motes are being dug.
• 1:00
  Speaker 1: Then we'll jump into the application layer, particularly a genetic AI and this whole workflow revolution people are talking about.
• 1:06
  Speaker 0: Yeah, that's a big one.
• 1:07
  Speaker 1: And finally, we have to tackle the non-technical side, which is often the hardest part.
• 1:12
  Speaker 1: Culture, risk, and how to adopt this responsibly.
• 1:16
  Speaker 0: Good plan.
• 1:17
  Speaker 0: So let's unpack that foundation first.
• 1:19
  Speaker 0: What's really fascinating, I think, is the high stakes game being played around compute power.
• 1:24
  Speaker 0: Definitely.
• 1:24
  Speaker 0: The whole idea of an AI moat is absolutely shifting.
• 1:28
  Speaker 0: It's not just about having the best model anymore.
• 1:30
  Speaker 0: It's becoming about who owns the entire pipeline.
• 1:34
  Speaker 0: The whole stack.
• 1:34
  Speaker 0: The whole vertical compute pipeline, yeah.
• 1:37
  Speaker 0: From the silicon right up to the user.
• 1:38
  Speaker 1: And maybe the biggest sign of this huge shift is seeing NVIDIA's grip on the GPU market starts to loosen a bit.
• 1:46
  Speaker 0: Right.
• 1:46
  Speaker 0: The standard for AI infrastructure for so long.
• 1:49
  Speaker 1: Exactly.
• 1:50
  Speaker 1: And Nehanth Maidu-Kalasetti pointed this out.
• 1:52
  Speaker 1: Oracle confirmed they're deploying what?
• 1:54
  Speaker 1: Fifty thousand AMD AI chips?
• 1:56
  Speaker 1: The MI-Four fifties?
• 1:57
  Speaker 1: Starting mid-twenty-twenty-six.
• 1:59
  Speaker 0: Fifty thousand?
• 2:01
  Speaker 0: That is a, well, a staggering number.
• 2:04
  Speaker 0: And it's not just about having a second source, right?
• 2:06
  Speaker 0: This looks like a really major alliance taking shape.
• 2:09
  Speaker 1: Oracle, AMD.
• 2:10
  Speaker 1: And
• 2:10
  Speaker 0: OpenAI.
• 2:11
  Speaker 0: They're making a huge bet that owning and optimizing these specific compute pipelines, well, that's a new oil field, basically.
• 2:19
  Speaker 0: It
• 2:19
  Speaker 1: feels like it.
• 2:20
  Speaker 1: I mean, that Oracle Cloud deal with OpenAI alone, potentially worth up to three hundred billion dollars.
• 2:25
  Speaker 1: Three
• 2:25
  Speaker 0: hundred billion, yeah.
• 2:27
  Speaker 0: That valuation tells you everything.
• 2:30
  Speaker 0: about the race cloud providers are in to lockdown supply and, you know, integrate hardware design right into their services.
• 2:36
  Speaker 1: It really is a full stack war now.
• 2:38
  Speaker 1: Makes it much tougher for newcomers trying to compete just on the model itself.
• 2:41
  Speaker 1: But,
• 2:41
  Speaker 0: okay, even if you figure out the chip supply, you immediately bump up against some pretty fundamental limits.
• 2:46
  Speaker 0: Walter Zuzavi highlighted this for major economies.
• 2:49
  Speaker 0: Like
• 2:49
  Speaker 1: the physical barriers.
• 2:50
  Speaker 0: Exactly.
• 2:51
  Speaker 0: In places like Germany, for instance, AI ambition is hitting a, well, a very hard wall.
• 2:56
  Speaker 0: Just basic data center power and capacity and
• 2:58
  Speaker 1: I saw those numbers the planned it load for twenty thirty something like three point seven gigawatts.
• 3:04
  Speaker 0: Yeah, but the projected demand could be closer to like, five point nine gigawatts.
• 3:09
  Speaker 0: Wow.
• 3:09
  Speaker 1: That gap was at two point two gigawatts.
• 3:11
  Speaker 1: That's a massive shortfall.
• 3:13
  Speaker 1: It definitely shifts capacity planning from just operations to a serious strategic risk at the board level.
• 3:20
  Speaker 0: Absolutely.
• 3:20
  Speaker 0: You can't really chase ambitious AI strategies, especially regulated ones, or build out sovereign clouds if you literally can't guarantee the power to run the hardware.
• 3:30
  Speaker 1: And that leads right into the next bottleneck, doesn't it?
• 3:33
  Speaker 1: connectivity.
• 3:34
  Speaker 0: It does.
• 3:35
  Speaker 0: Alan Weckel noted that these huge AI networks often stumble not because the GPUs aren't fast enough, but because of the underlying network fabric.
• 3:42
  Speaker 1: Ah, the fabric.
• 3:43
  Speaker 1: So for people maybe less deep in networking, we're talking the high-speed connections linking all those thousands of GPUs.
• 3:50
  Speaker 0: Precisely.
• 3:51
  Speaker 0: Option non-blocking Ethernet designs.
• 3:53
  Speaker 0: If that fabric isn't absolutely perfect, the second you try to distribute a really large training job, performance just tanks.
• 3:59
  Speaker 1: Right.
• 3:59
  Speaker 1: It grinds to a halt.
• 4:00
  Speaker 0: Yeah.
• 4:01
  Speaker 0: The fabric is like the central nervous system.
• 4:04
  Speaker 0: It has to handle enormous amounts of data, all moving in parallel, no collisions.
• 4:09
  Speaker 0: These non-blocking designs are basically non-negotiable now for the kind of scale you need in, say, Neo or sovereign clouds.
• 4:16
  Speaker 0: Thanks, us.
• 4:16
  Speaker 0: And Rick Moore mentioned solutions like Lumen's Wavelength Rapid Routes offering, you know, four hundred GBPs dedicated links.
• 4:23
  Speaker 0: They're specifically built to kill those bottlenecks, especially when your architecture is spread out geographically.
• 4:28
  Speaker 1: Okay, so... That focus on connectivity, performance at scale, it ties directly into what Armand Ruiz observed, doesn't it?
• 4:37
  Speaker 1: This shift in industry focus away from the super expensive training phase.
• 4:41
  Speaker 1: Torch
• 4:41
  Speaker 0: inference, yeah.
• 4:42
  Speaker 0: Inference.
• 4:43
  Speaker 0: Model deployment, execution.
• 4:45
  Speaker 0: That's what dominates the long-term AI workload.
• 4:47
  Speaker 0: Training is that big upfront cost, sure, but... inference is the ongoing operational cost.
• 4:52
  Speaker 1: So the drive for efficiency there, especially at the edge where the data often is, that's where the real ROI is going to be.
• 4:58
  Speaker 1: That's
• 4:58
  Speaker 0: the thinking.
• 4:59
  Speaker 0: It's all about efficient execution for lasting value.
• 5:02
  Speaker 1: I do wonder though, if the focus shifts so hard to inference efficiency, does that mean training budgets actually shrink eventually?
• 5:10
  Speaker 1: Or does the complexity just keep pushing the training costs up anyway?
• 5:13
  Speaker 0: That's the million dollar question, isn't it?
• 5:14
  Speaker 0: Or maybe the billion dollar question now.
• 5:16
  Speaker 0: But the immediate goal is definitely ROI, getting value out of these things.
• 5:20
  Speaker 0: And that sets us up perfectly for our next theme, the agentic enterprise.
• 5:24
  Speaker 0: Right.
• 5:24
  Speaker 0: Because that need for efficient inference is exactly what fuels the drive for these more autonomous systems, systems that can actually handle multi-step workflows.
• 5:34
  Speaker 1: We are definitely seeing that shift.
• 5:36
  Speaker 1: Moving beyond just static tools to systems that can, you know, reason, plan, act on their own, Tony Moroni laid out four ways these agents could unlock serious enterprise value.
• 5:47
  Speaker 0: Okay, like what?
• 5:48
  Speaker 1: Well, expanding automation, obviously.
• 5:50
  Speaker 1: Having systems available to it over for four or seven.
• 5:52
  Speaker 1: building more organizational resilience, and, maybe most importantly, turning all the dormant knowledge companies have directly into action.
• 6:01
  Speaker 1: Verifiable
• 6:01
  Speaker 0: action.
• 6:02
  Speaker 0: And to do that, you need the right technical underpinning.
• 6:05
  Speaker 0: That's why Eduardo Ordex called agentic rag plus memory the missing link.
• 6:10
  Speaker 1: OK, so rag, retrieval augmented generation, most people know that.
• 6:13
  Speaker 1: It's how you ground the model in your own documents, right?
• 6:15
  Speaker 1: Yeah.
• 6:15
  Speaker 1: Stop from just making things.
• 6:16
  Speaker 1: Exactly.
• 6:17
  Speaker 0: Stops the hallucinations or tries to.
• 6:19
  Speaker 0: But agentic rag plus memory adds three really key things on top.
• 6:25
  Speaker 0: An orchestrator, specific memory functions, and these reasoning loops, like react.
• 6:30
  Speaker 0: OK.
• 6:31
  Speaker 0: which lets the system actually plan dynamically.
• 6:34
  Speaker 0: Not just grab a chunk of text, but break down a complex task, search iteratively, synthesize info over multiple steps.
• 6:40
  Speaker 0: You know, solve real messy enterprise problems.
• 6:43
  Speaker 1: So instead of just asking what was Q three revenue and getting a number, you could analyze Q three revenue against Q four cost projections and suggest three ways to cut costs.
• 6:54
  Speaker 1: Something that needs planning.
• 6:56
  Speaker 0: That's the idea.
• 6:56
  Speaker 0: That's the practical leap forward.
• 6:58
  Speaker 0: And we're seeing this pop up in actual commerce.
• 7:00
  Speaker 0: Hugo Caterino mentioned how in agentic search, these assistants are trying to unify discovery, trust, and even settlement.
• 7:06
  Speaker 0: So search relevance suddenly has to include checking price, stock, whether the payment method works.
• 7:13
  Speaker 0: because if the agent suggests buying it, it better be immediately doable.
• 7:17
  Speaker 1: That really raises the stakes, doesn't it?
• 7:19
  Speaker 1: The liability and the expectations on the AI in a transaction?
• 7:22
  Speaker 0: It absolutely does, but it also creates opportunities.
• 7:25
  Speaker 0: Anna Lerner-Nesbitt was talking about the Walmart and open AI shopping experience.
• 7:30
  Speaker 0: Oh, yes,
• 7:30
  Speaker 1: all
• 7:30
  Speaker 0: that.
• 7:31
  Speaker 0: And she raised a really interesting point.
• 7:33
  Speaker 0: Agents could be a huge opportunity to bake in sustainable choices from the start.
• 7:38
  Speaker 1: How so?
• 7:39
  Speaker 0: Well, they could act like automated choice architects, right?
• 7:43
  Speaker 0: Prioritizing efficiency, maybe suggesting lower waste options, guiding users towards more sustainable choices across the value chain, potentially without needing constant human intervention.
• 7:54
  Speaker 1: That's a powerful concept.
• 7:55
  Speaker 1: Using automation to nudge behavior towards positive outcomes.
• 7:59
  Speaker 1: But we have to get real about deployment.
• 8:01
  Speaker 1: It's proving really tough.
• 8:02
  Speaker 0: Oh, absolutely.
• 8:04
  Speaker 0: Virtrungland.DA cited an HBR report pointing out that most companies are really struggling to scale AI beyond the pussies
• 8:10
  Speaker 1: stage.
• 8:10
  Speaker 1: Why is that?
• 8:11
  Speaker 0: Performance drops, unpredictable costs, and big security gaps are the main culprits.
• 8:16
  Speaker 1: Okay, so what's the fix?
• 8:17
  Speaker 0: The emerging consensus seems to be around something called generative operations, or GenOps.
• 8:22
  Speaker 0: Think MLO's rigor, but specifically tailored for these generative models.
• 8:27
  Speaker 1: So a structured way to manage deployment.
• 8:29
  Speaker 0: Exactly.
• 8:30
  Speaker 0: Automated monitoring, drift detection, cost analysis, governance checks baked in before, during, and after deployment.
• 8:36
  Speaker 0: Without that kind of operational discipline, these agents just won't scale responsibly or maybe even safely.
• 8:42
  Speaker 1: Which brings us right to our last theme.
• 8:44
  Speaker 1: Culture, risk, and responsible adoption.
• 8:47
  Speaker 0: Yeah.
• 8:48
  Speaker 1: Because the best infrastructure and the smartest agents, well, they're kind of useless if the organization isn't ready.
• 8:54
  Speaker 0: That really is the heart of it, isn't it?
• 8:56
  Speaker 0: The biggest hurdle often isn't technical at all.
• 8:58
  Speaker 0: There is an IBM survey of execs.
• 9:01
  Speaker 1: Yeah.
• 9:01
  Speaker 0: Sixty-seven percent expect agents to act independently by twenty-twenty-seven.
• 9:05
  Speaker 0: Huge number.
• 9:07
  Speaker 0: But the main barrier they see, not the tech, it's
• 9:10
  Speaker 1: people,
• 9:11
  Speaker 0: mindset, organizational readiness.
• 9:13
  Speaker 1: It always comes back to the people.
• 9:15
  Speaker 1: Pretty, WaterD made that point really clearly.
• 9:18
  Speaker 1: Sustainable AI impact comes from prepared people, not just powerful platforms.
• 9:22
  Speaker 0: Right.
• 9:23
  Speaker 1: And the stats back it up.
• 9:24
  Speaker 1: Something like seventy-eight percent of global orgs are using AI now, but fewer than one percent are actually considered fully mature in how they use it.
• 9:32
  Speaker 0: And you see that maturity gap reflected nationally too.
• 9:35
  Speaker 0: Donor Mustafa highlighted the situation in Sweden pretty striking.
• 9:39
  Speaker 0: Seventy-seven percent of Swedes using AI at work have had zero formal education or training on it.
• 9:46
  Speaker 1: seventy seven percent.
• 9:47
  Speaker 1: wow yeah that's.
• 9:49
  Speaker 1: that's not an employee problem that's a leadership failure.
• 9:51
  Speaker 1: surely absolutely.
• 9:52
  Speaker 0: when adoption happens bottom up like that.
• 9:54
  Speaker 0: and leadership doesn't provide guardrails doesn't provide training that's a massive risk
• 9:59
  Speaker 1: which ties into navine balani's reminder that ai really needs to be treated as an engineering discipline not just you know a bunch of cool experiments
• 10:07
  Speaker 0: totally.
• 10:07
  Speaker 0: The differentiator isn't just having the latest LLM or a co-pilot.
• 10:11
  Speaker 0: It's the engineering mindset behind how you deploy, manage, and integrate these tools into critical workflows.
• 10:17
  Speaker 0: Rigor matters.
• 10:18
  Speaker 1: And that rigor needs to extend down to the end user too.
• 10:22
  Speaker 1: Daniel Paul has some really practical advice on getting more out of existing models, like GPT-V.
• 10:27
  Speaker 1: Yeah.
• 10:28
  Speaker 1: Using its different modes properly.
• 10:30
  Speaker 1: like instant mode for quick ideas, thinking mode for strategy, pro mode for deep, precise work, auto mode, if you're not sure, moving beyond just tossing random prompts at it.
• 10:40
  Speaker 1: That
• 10:40
  Speaker 0: makes sense.
• 10:41
  Speaker 0: Getting intentional about how you interact with it, because that attention to detail is critical, especially when you factor in the ethical stuff and the risks involved.
• 10:48
  Speaker 1: Definitely.
• 10:48
  Speaker 1: And one of the most sobering things we saw came from Dr.
• 10:51
  Speaker 1: Hans Ruznick citing that Harvard study.
• 10:53
  Speaker 1: The
• 10:54
  Speaker 0: weird bias one.
• 10:55
  Speaker 1: Exactly.
• 10:55
  Speaker 1: Showing how LLMs are heavily biased towards Western educated industrial rich and democratic perspectives.
• 11:02
  Speaker 0: That's a fundamental challenge.
• 11:03
  Speaker 0: If the training data is skewed towards one narrow cultural viewpoint, how fair or applicable are these models globally?
• 11:11
  Speaker 0: It raises huge questions about governance, trust, especially for decisions impacting devotes groups.
• 11:17
  Speaker 0: For sure.
• 11:17
  Speaker 1: And even when the output sounds right, Kassi Kazarkov warns about the fluency heuristic.
• 11:23
  Speaker 0: Ah, yeah, the tendency to trust something just because it sounds smooth and confident.
• 11:26
  Speaker 1: Exactly.
• 11:27
  Speaker 1: It's a dangerous psychological trap.
• 11:29
  Speaker 1: The AI sounds so sure of itself that humans might skip the crucial step of actually verifying the output.
• 11:36
  Speaker 0: Which can lead to really costly mistakes, right?
• 11:38
  Speaker 0: Because the model will hallucinate or make subtle errors eventually.
• 11:41
  Speaker 0: You almost have to default to assuming it's wrong until you prove it right.
• 11:45
  Speaker 1: And the final piece of the risk puzzle.
• 11:47
  Speaker 1: Security.
• 11:48
  Speaker 1: with these autonomous agents.
• 11:50
  Speaker 1: Hussaini Savawala pointed out that AI coding agents, well, they're not just speeding up development.
• 11:56
  Speaker 0: They're speeding up vulnerability creation, too.
• 11:58
  Speaker 1: Unfortunately, yeah.
• 11:59
  Speaker 1: Replicating security flaws at enterprise scale incredibly quickly.
• 12:04
  Speaker 0: So what's the answer there?
• 12:06
  Speaker 1: It seems to be what they call data-first agentic security.
• 12:09
  Speaker 1: Basically, build security in from the start.
• 12:12
  Speaker 1: prevent sensitive data from even getting into the model's context if possible.
• 12:16
  Speaker 1: Don't try to just patch things later.
• 12:18
  Speaker 0: Makes sense.
• 12:19
  Speaker 0: Security as a foundation, not an afterthought.
• 12:22
  Speaker 1: So pulling it all together from the insights these past couple of weeks.
• 12:25
  Speaker 1: Yeah.
• 12:26
  Speaker 1: What's the big picture?
• 12:28
  Speaker 1: It really feels like the market's moving past the initial hype, getting down to the foundations.
• 12:32
  Speaker 0: Definitely maturing.
• 12:34
  Speaker 1: Yeah.
• 12:34
  Speaker 1: And that demands serious strategic investment in infrastructure, whether it's new chip deals or perfecting that network fabric.
• 12:42
  Speaker 1: But maybe even more crucial is the shift needed in company culture and engineering discipline to actually manage these powerful new agentic systems properly.
• 12:51
  Speaker 0: And Kira Dotson put it pretty bluntly.
• 12:53
  Speaker 0: Yeah.
• 12:54
  Speaker 0: The era of AI for the sake of AI is over.
• 12:57
  Speaker 0: Right.
• 12:57
  Speaker 0: So the question every organization needs to ask itself now is, are we still chasing vanity metrics just looking busy with AI?
• 13:04
  Speaker 0: Or are we truly focused on delivering measurable enterprise grade value that's going to last beyond the current hype cycle and actually reshape how we operate?
• 13:13
  Speaker 1: That really is the key differentiator now.
• 13:15
  Speaker 1: Focus on real value.
• 13:16
  Speaker 0: Exactly.
• 13:17
  Speaker 0: Are you building something that endures or just riding the wave?
• 13:20
  Speaker 1: Well said.
• 13:21
  Speaker 1: If you enjoyed this deep dive, new episodes drop every two weeks.
• 13:25
  Speaker 1: Also check out our other editions on ICT and tech.
• 13:28
  Speaker 1: Digital products and services, cloud, sustainability in green ICT, defense tech and health tech.
• 13:34
  Speaker 0: Thanks for tuning in.
• 13:35
  Speaker 0: We really encourage you to subscribe so you don't miss our continued deep dives and to the insights that matter most for your strategy.