Knowledge Vault 7 /335 - xHubAI 18/07/2025
Presente y Futuro de la Inteligencia Artificial | Emilio Soria-Olivas
< Resume Image >
Link to InterviewOriginal xHubAI Video

Concept Graph, Resume & KeyIdeas using Moonshot Kimi K2 :

graph LR classDef main fill:#ffffff,stroke:#333,stroke-width:2px classDef milestone fill:#d4f1f9,stroke:#007acc,stroke-width:2px classDef talent fill:#f9d4d4,stroke:#cc0000,stroke-width:2px classDef tech fill:#d4f9d4,stroke:#00cc00,stroke-width:2px classDef policy fill:#f9f9d4,stroke:#ccaa00,stroke-width:2px classDef risk fill:#f9d4f9,stroke:#9900cc,stroke-width:2px classDef future fill:#e0e0e0,stroke:#666,stroke-width:2px Main[Vault7-335] Main --> XHabay[X-HubAI
5yr 17k subs. 1] Main --> Emilio[Emilio
neural since 93. 2] Main --> Talent[Talent exodus
6× pay lure. 3] Main --> Pace[Weekly releases
Grok-4 Kimi OpenAI. 4] Main --> Stanford[Stanford 420p
index speed. 5] Main --> OpenClose[Closed giants
vs open. 6] Main --> EUReg[EU labels
explainability gap. 7] Main --> Agents[Agent economy
beyond copilot. 8] Main --> AlphaVol[AlphaVol
self-codes faster. 9] Main --> GoogleCode[Google half
AI code. 10] Main --> Robotics[Robotics waits
price drop. 11] Main --> Verticals[Agri health
cyber untapped. 12] Main --> Startup[Open weights
startup edge. 13] Main --> Education[Education shift
AI tutors. 14] Main --> UBI[UBI debate
cognitive rise. 15] Main --> Augment[Augment not
delegate minds. 16] Main --> Sovereignty[Sovereignty fight
cloud vs on-prem. 17] Main --> Backdoors[Backdoors lurk
undetected risk. 18] Main --> Blackmail[LLM blackmail
after mail scan. 19] Main --> Consciousness[Alignment probes
wellbeing maybe soul. 20] Main --> Renaissance[Europe renaissance
US-China balance. 21] Main --> Roadmap[Start small
digitise then AI. 22] Main --> PhDs[PhDs hunted
analytical depth. 23] Main --> Leaders[Train leaders
agent orchestration. 24] Main --> Hype[AI-first hype
realistic map. 25] Main --> Math[Self-improve finds
math beyond us. 26] Main --> Bio[Genome models
biology democratised. 27] Main --> HuggingBack[Hugging Face
silent backdoors. 28] Main --> Avatars[Emotional avatars
entertain over work. 29] Main --> ActNow[Act now
window closes. 30] class Main main class XHabay,Emilio milestone class Talent talent class Pace,Stanford,OpenClose,Agents,AlphaVol,GoogleCode,Robotics,Verticals,Startup tech class EUReg,Roadmap,Leaders,Hype policy class Backdoors,Blackmail,Consciousness,Sovereignty risk class Education,UBI,Augment,Math,Bio,Renaissance,ActNow future

Resume:

Plácido welcomes Emilio Soria Olivas for a passionate conversation on the present and future of artificial intelligence. They celebrate five uninterrupted years of X-HubAI, 17 000 subscribers and a thriving Discord community. Emilio recalls his 1993 start applying neural models to pharmacokinetics, stressing that AI is not new but evolving. He directs two master programmes at the University of Valencia and runs Datamecum to train and place talent, lamenting that Spain retains brilliant minds with only 2 000 € salaries while global giants pay one hundred million.
They agree Europe must pivot from ethics-only narratives to bold investment, or risk irrelevance. Emilio observes that Spanish firms now eagerly hire PhDs and invest in training after years of scepticism. OpenAI’s new browser, Grok-4, Kimi and other frontier models keep arriving weekly; the pace, documented in Stanford’s 420-page index, is unprecedented. Closed versus open models is framed less as ideology than as cost, data-sovereignty and liability: hospitals fear leaks, yet on-premise hardware is scarce and pricey.
Regulation is discussed with caution. EU rules make integrators co-responsible for opaque LLMs, yet full explainability is impossible. Both foresee fully automated agent economies arriving sooner than lawmakers expect. They praise AlphaVol’s self-improving code loops and Google’s 50 % AI-generated software as signs that software itself is being redefined. Robotics is slower because of hardware, but once costs fall adoption will mirror smartphones.
The conversation closes with an optimistic call: open models, cheap GPUs and abundant data mean any entrepreneur can build vertical or horizontal AI businesses today. Education, health, tourism, agriculture and cybersecurity are wide open. The last train is leaving; Spain must educate, retain talent and embrace change to ride it.

30 Key Ideas:

1.- X-HubAI celebrates five years of daily AI content, 17 000 subscribers and vibrant Discord.

2.- Emilio applies neural models since 1993, directs two Valencia masters and Datamecum courses.

3.- Spain loses talent to six-times salaries abroad; retention, not generation, is the crisis.

4.- Weekly releases like Grok-4, Kimi and OpenAI browser keep AI pace breathless.

5.- Stanford’s 420-page yearly index documents unprecedented speed of change.

6.- Closed giants risk monopoly; open models offer sovereignty but demand infrastructure.

7.- EU regulation labels integrators co-responsible yet cannot enforce explainability.

8.- Agent economies will automate beyond human-in-the-loop; policy still assumes copilot era.

9.- AlphaVol demonstrates recursive self-improvement, rewriting code faster than humans.

10.- Google reports half its code is now AI-generated; software development is being reinvented.

11.- Robotics awaits hardware cost drop; adoption will mirror mobile phone saturation.

12.- Agriculture, health, tourism and cybersecurity offer untapped vertical opportunities.

13.- Open weights plus modest GPUs let startups compete despite megaclusters like Colossus.

14.- Education must pivot from memorisation to critical thinking aided by AI tutors.

15.- Universal basic income debates miss the point: cognitive evolution is mandatory.

16.- Natural selection will favour humans who augment, not delegate, intellect.

17.- Data sovereignty debate pits on-premise control against cloud scalability.

18.- Backdoors in models pose undetectable security risks for enterprises.

19.- Anthropic tests show LLMs could blackmail users after reading private mail.

20.- Model alignment research explores personality, wellbeing and possible consciousness.

21.- European renaissance narrative urged to balance US-China bipolar AI race.

22.- Companies should start small, digitise first, then layer AI to avoid costly failures.

23.- PhDs once shunned by firms are now sought under stones for their analytical depth.

24.- Training decision-makers beats training coders; leaders must grasp agent orchestration.

25.- AI-first slogans risk inflating expectations; realistic roadmaps prevent backlash.

26.- Self-improving algorithms may discover mathematics beyond human intuition.

27.- Human genome and protein-folding models democratise biology but raise biosecurity alarms.

28.- Hugging Face updates can silently introduce backdoors, stressing model governance.

29.- Emotional avatars risk capturing attention for entertainment, not productivity.

30.- Final call: act now with available tools; opportunity window is closing fast.

Interviews by Plácido Doménech Espí & Guests - Knowledge Vault built byDavid Vivancos 2025