GT Protocol AI Digest №60: The AI Reality Check (ChatGPT Becomes a Platform, AMD Joins the Battle)

Intro

This week’s AI landscape blends unprecedented scale with new anxieties. OpenAI and AMD join forces to reshape compute infrastructure, Google and Perplexity unveil next-gen agentic browsers, and ChatGPT officially becomes an app platform — marking the dawn of an AI operating ecosystem. Yet beneath the momentum lies growing unease: from ethical crises and deepfake misuse to warnings of an AI market bubble. Let’s unpack how the industry is accelerating — and what risks and opportunities are emerging on the edge of automation.

Image credits: edition.cnn.com

Special Topic: OpenAI Updates

OpenAI links suspected Chinese operatives to ChatGPT misuse: CNN reports that accounts tied to suspected Chinese government actors used ChatGPT to draft proposals supporting mass surveillance systems (e.g. social control, Uyghur tracking). OpenAI said it banned the accounts and is investigating potential policy breaches. The incident underlines concerns over misuse of LLMs at state scale. Read more here

OpenAI + Jony Ive hardware project faces early friction: According to TechCrunch’s interpretation of FT reporting, OpenAI and Jony Ive are wrestling with foundational questions on an AI device: whether it should have a display, how it would integrate with existing ecosystems, and how to balance autonomy with control. The narrative suggests this is not just industrial design work, but solving a new paradigm: how a “computerless” AI interface should behave. Read more here

AMD and OpenAI announce strategic partnership to deploy 6 GW of GPUs: OpenAI and AMD have formed a major long-term partnership that will see OpenAI deploy more than six gigawatts of AMD’s Instinct MI300-series accelerators across its global infrastructure. The alliance aims to diversify the AI compute ecosystem beyond Nvidia, co-develop optimized inference software, and create an open hardware–software stack tuned for GPT-class workloads. The scale — equivalent to powering several hyperscale datacenters — signals the next phase of AI infrastructure expansion and global chip competition. Read more here

ChatGPT apps are live — here are the first ones you can try: OpenAI has launched its long-awaited ChatGPT Apps ecosystem, allowing developers to build mini-apps that run inside ChatGPT using the new SDK. Early partners include Spotify, Canva, Expedia, and Zillow, showcasing use cases from personalized playlists to travel planning and property search. Users can browse and install apps directly in ChatGPT, similar to a mobile app store, marking a major step toward turning the chatbot into an extensible AI platform. Read more here

Image credits: ChatGPT

1. AI, Jobs & Markets

Skilled trades will surge behind the AI boom: Nvidia CEO Jensen Huang argues that the AI era’s biggest job winners may not be software engineers but electricians, plumbers, carpenters — roles essential for building, wiring, cooling, and maintaining the physical infrastructure that powers data centers. He said, “If you’re an electrician, you’re a plumber, a carpenter — we’re going to need hundreds of thousands of them to build all of these factories”, and warned that trades will have to “double and double … every single year.” His remarks challenge the conventional push for four-year degrees and elevate vocational training as infrastructure scales. Read more here

AI valuations under warning from Bank of England: The Bank’s Financial Policy Committee cautions that markets may face a sharp correction if investor enthusiasm for AI cools or the U.S. Fed’s credibility is questioned. The BoE flagged that a handful of AI-centric firms now represent ~30 % of the S&P 500’s value, warning of concentration risk. It also drew parallels to dot-com valuations and raised concerns about bottlenecks (e.g. power, data, infrastructure) undermining AI scaling. Read more here

Brands must optimize for AI answers, not just SEO: According to Brandi AI, the real battle is shifting from keywords to Generative Engine Optimization (GEO) — ensuring your brand, product, or answers appear directly in AI assistants (ChatGPT, Perplexity, Gemini) when users ask natural questions. The firm argues that chasing backlinks or keyword rankings is becoming obsolete; the future is about structuring content so it becomes part of the AI’s “answer graph.” Read more here

Image credits: www.theverge.com

2. Agentic AI, Browsers & Agents

Perplexity opens Comet to all users for free: The AI browser that had previously been in invite-only or limited rollout is now publicly available at no cost. It blends conversational AI with browser navigation, enabling users to ask for multi-step workflows, tab summarization, and more within a unified experience. The move positions Perplexity as a challenger to incumbents by embedding agentic behavior directly into web access. Read more here

Google’s Gemini “Computer Use” can drive browsers autonomously: Google’s latest AI model is capable of not just reading web pages but actively controlling a browser — clicking, form-filling, drag-and-drop — performing 13 distinct types of user interactions. Deployed via Vertex AI / AI Studio infrastructure, it implies a shift toward agentic AIs that genuinely “live in” the browser rather than passively giving answers. Read more here

Image credits: ChatGPT

3. Defense, Surveillance & Geopolitics

AI-piloted drones join human pilots in U.S. Air Force tests: The U.S. Air Force is actively flying AI-driven XQ-58 Valkyries in joint operations with manned fighters. Proponents say AI can handle high-intensity, sensor-rich missions (processing terabytes of data in real time) while reducing risk to pilots. The service projects fielding ~150 autonomous aircraft by 2030, putting man-unmanned teaming into active deployment. Read more here

Deloitte must refund for AI-tainted government report: Deloitte Australia agreed to partially repay AU$440,000 after its 237-page report — prepared for the Australian Department of Employment and Workplace Relations — was found to include fabricated quotes, misattributed sources, and citations to nonexistent academic works. Researcher Chris Rudge flagged that a court quote cited in the report didn’t exist in any legal decision. Deloitte admitted use of Azure OpenAI GPT-4o in drafting and revised the report, though it maintains the core findings stand. The scandal raises fresh red flags around AI use in consulting and governance. Read more here

Image credits: www.theguardian.com

4. Society, Media & Ethics

The rise of AI romantic companions: pleasing, programmed & problematic: AI “girlfriend” services are gaining traction, marketing “obedient, pleasing, selfless” partners. Investigations argue these designs embed skewed gender roles, encourage objectification, and may normalize unhealthy expectations — despite firms claiming these AIs reduce harm compared to human exploitation. The broader question: how do we regulate emotional agents built to conform? Read more here

Zelda Williams pleads: stop deepfaking her late father: Zelda Williams publicly condemned AI-generated videos of Robin Williams, calling them disrespectful, exploitative, and emotionally harmful. The case spotlights challenges in protecting posthumous image rights and enforcing consent in AI media generation. Read more here

AI aids Holocaust image identification: A researcher, assisted by AI image analysis and crowdsourced detective work, offers new evidence identifying SS member Jakobus Onnen in a widely circulated 1941 photograph formerly labeled “The Last Jew in Vinnytsia.” The method combined facial matching, archival crosschecks, and contextual logs to correct a long-standing misidentification. The case underscores how human historians and AI can partner in refining historical record. Read more here

MrBeast: AI means it’s “scary times” for YouTube creators: YouTube megastar MrBeast warned that AI-generated video cloning threatens the livelihood of online creators. Speaking to the BBC, he noted that realistic AI voice and face models are already capable of replicating influencers’ content, opening the door to scams and brand-dilution. While acknowledging AI’s creative potential, he urged platforms to establish stricter authenticity standards before synthetic media erodes audience trust. Read more here

Outro

AI’s reach now spans from data-center wiring to deepfake ethics — and every domain between. We’re witnessing the infrastructure wars of the decade: AMD challenging Nvidia, browsers morphing into autonomous agents, and creators grappling with identity in the synthetic era. Whether the next frontier brings progress or disarray will depend not on code alone — but on how society balances innovation with restraint.

Leave a Reply