Stay ahead in the AI race — follow Global Post Headline for in-depth coverage, expert insights, and the stories shaping tomorrow’s intelligence.

The AI Race: Big Steps, Big Money, but Still Miles from the Finish Line

August 10, 2025

When Sam Altman, CEO of OpenAI, described the company’s latest GPT-5 upgrade as a “significant step on the path to AGI” but “missing something quite important”, it was a bit like saying your new car is faster, shinier, and smarter—yet still can’t drive itself without you in the seat.

That “finish line” Altman referred to is AGI—artificial general intelligence. Think of it as AI that can do almost everything a human can do, from writing reports and designing clothes to running a business meeting or even fixing a broken washing machine… well, virtually, at least.

But here’s the twist: despite billions of dollars and brainpower from companies like OpenAI, Google, Meta, and Chinese firms such as DeepSeek, no one’s quite there yet.

Where We Are Now

Today’s AI tools—ChatGPT, Claude, Gemini, Llama, DeepSeek—are already changing how we work, learn, and create. Businesses use them to write marketing campaigns, detect fraud, and automate customer service. Students use them for research. Filmmakers use them to storyboard scenes.

Yet, as Altman himself admitted, GPT-5 still can’t “continuously learn” after launch—meaning it doesn’t improve itself like humans do through daily experience.

Tech analyst Benedict Evans puts it bluntly: we don’t even fully understand why these models work so well. It’s a little like building a rocket without complete knowledge of gravity: you can keep making the rocket bigger, but you might not get to the moon just from scale alone.

The Players and Their Strategies

OpenAI – Released “open weight” models (gpt-oss-120b and gpt-oss-20b-two) so developers can download and customise powerful models while keeping some internal opacity compared with full open source.
Meta (Facebook) – Publishes Llama models for wide use and fine-tuning, arguing this democratizes AI access while also cautioning around more advanced models.
Google DeepMind – Focuses on “world models” (e.g., Genie 3) that simulate realistic environments to train AI for robotics and agent-like behaviour.
DeepSeek and other Chinese firms – Have rapidly produced capable models, with real-world corporate customers like Saudi Aramco applying them in mission-critical systems.

This competition is not just about technology—it's economic and geopolitical. Microsoft’s Brad Smith argued before the U.S. Senate that whichever country gets its AI adopted globally first gains a huge advantage, echoing lessons from previous tech wars like Huawei and 5G.

The Business Behind the Buzz

Even without AGI, AI is a huge commercial success. OpenAI’s annual recurring revenue jumped to around $13 billion, with projections that it could pass $20 billion within a short span.

Industry spending is enormous: major U.S. tech firms—Alphabet, Meta, Microsoft, Amazon—are expected to spend a collective hundreds of billions on AI this year, far outstripping other major budgets such as EU defence spending.

If OpenAI went public through an IPO, it could become one of the largest tech listings ever—assuming market expectations align with real product utility. But as with any hype-driven sector, big valuations come with big scrutiny.

New Products and Everyday Examples

Small business: A corner shop uses a GPT-powered tool to analyse sales data, recommend what products to stock next week, and auto-generate local social media posts that drive customers through the door.

Education: Teachers deploy customised AI tutors that adapt to each student’s learning pace, offering extra practice where needed and freeing teachers for personalised coaching.

Healthcare: Multimodal AI helps triage cases by reading scans, checking medical history, and flagging potentially urgent issues so clinicians can prioritise patients faster.

With OpenAI’s “open weight” moves and Meta’s Llama, developers can now build specialised agents—like a landlord-friendly legal document summariser or a travel planner that builds a full itinerary tailored to mobility and budget constraints.

The Future: AGI or Just Smarter Tools?

Experts don’t agree on what AGI even means. Some, like venture investor Aaron Rosenberg, argue a narrower interpretation—AI that reaches 80th-percentile human performance across most digital tasks—could be realistic within five to ten years. Others say the goalposts will keep shifting.

Superintelligence—a system smarter than humans across almost all cognitive tasks—remains speculative and distant. Right now, competitors are jockeying to have their models become the global default, because broad adoption is a moat that’s hard to cross.

Risks in the Rush

Free and powerful models raise serious concerns. Some experts fear they could be twisted into tools for harmful uses—everything from social-engineering campaigns to biological or cyber threats. There are also more immediate issues: biased outputs, misinformation, and failures in critical contexts.

Researchers and ethicists warn against confusing marketing language with technical reality. As David Bader notes, improvements in reasoning and multimodal understanding are real—but they are not proof that superintelligence is imminent.

Final Takeaway

We are in the middle of a fast-moving, global AI race. Recent advances—GPT-5, Genie 3, DeepSeek R1—are impressive but are steps, not leaps. For most people and companies, AI’s immediate value will be in augmenting human tasks, not replacing humans wholesale.

The winners will be those who ship reliable, widely adopted products that solve real problems. Whether the next major breakthrough comes from Silicon Valley, Beijing, or somewhere unexpected, the race will keep accelerating—and the stakes are enormous.