The race to artificial general intelligence (AGI) dominance is getting fiercely competitive. Former Google CEO Eric Schmidt recently predicted that the leaders in the generative AI revolution will emerge within the next few monthsas innovation accelerates at an unprecedented pace.
“The winners in the generative revolution—code, videos, text, everything—are being determined in the next six to 12 months. The growth rate is quadrupling every six months, and once the slope is set, it’s incredibly hard for competitors to catch up,” Schmidt said, adding that this critical window would leave slower competitors struggling to stay relevant.
Citing unprecedented scaling and the absence of limiting factors in current technology, he said that AI models will be 50 to 100 times more powerful in five years.
“In five years, we’ll see two or three more turns of the crank of these large models. Each turn could result in systems 50 to 100 times more powerful,” Schmidt further said. According to him, rapid scaling is a critical factor driving AGI development.
“Who doesn’t want to be the first on that mountain?” said NVIDIA chief Jensen Huang at the podcast No Priors, pointing out that the race to reach AGI is getting fierce, with major players like OpenAI, Anthropic, xAI, alongside Google, Meta, and Microsoft, all competing to lead.
“The prize for reinventing intelligence altogether…it’s too consequential to not attempt it,” he added, noting that scaling laws and massive computational advances are crucial.
“We’re close to AGI…but even if we could argue about whether it is really general intelligence, just getting close to it is going to be a miracle…Everything is going to be hard…but nothing is impossible,” said Huang.
Who Will Lead AGI’s Future?
OpenAI believes it is ahead of the pack with its Strawberry (o1), which uses synthetic data generated by models like GPT-5 to create a recursive cycle of improvement. Each successive GPT version benefits from the high-quality data of its predecessor, ensuring continual enhancements in capabilities.
While OpenAI chief Sam Altman dismissed concerns about scaling limits, asserting, “There is no wall,” former OpenAI chief scientist Ilya Sutskever emphasised the importance of scaling strategically. “Scaling the right thing matters more now than ever,” Sutskever added.
OpenAI’s approach reflects confidence but raises questions about long-term viability in the face of potential diminishing returns.
Meta, on the other hand, is forging its path to AGI by moving beyond traditional LLMs, focusing on human-like reasoning through autonomous machine intelligence (AMI). Under chief scientist Yann LeCun’s leadership, Meta is developing systems like Layer Skip and V-JEPA to enhance machines’ ability to reason and interact with the world, aka ‘world models’, which Jurgen Schmidhuber claims are a rehash of his work.
Meta has also integrated self-supervised learning to power its upcoming Llama 4 and recently introduced tools like the Self-Taught Evaluator, which uses chain-of-thought reasoning to break complex problems into smaller steps. These efforts signify Meta’s commitment to creating adaptable, human-aligned AI systems.
Meanwhile, Anthropic is striving to find its breakthrough moment, balancing inspiration from OpenAI with its search for unique scaling methods.
While employing synthetic data and reinforcement learning for its Claude series, including the upcoming Claude 3.5 Opus and Claude 4, CEO Dario Amodei remains sceptical of current approaches. “Even if there’s no problem with data, as we start to scale models up they just stop getting better,” he warned.
Anthropic is experimenting with techniques like dictionary learning to identify patterns in neuron activations, aiming to address LLM limitations. The company’s focus on finding a ‘new architecture’ highlights its intent to redefine the boundaries of scaling.
Google DeepMind is blending scaling with architectural innovation, betting on multimodal and neuro-symbolic AI to propel it towards AGI.
Despite high expectations for its Gemini series, the models have underperformed, highlighting the challenges in translating scaling into meaningful performance gains. DeepMind CEO Demis Hassabis insisted on the need for new algorithms, stressing, “Half our efforts have to do with inventing the next architectures and the next algorithms.”
Google DeepMind is also investing in RL agents, inspired by the success of AlphaGo and AlphaZero, and exploring real-world simulations to improve understanding of complex systems. Neuro-symbolic models like AlphaProof and AlphaGeometry reflect its strategic pivot to prevent generative AI from hitting a performance ceiling.
AGI of Thrones
Elon Musk’s xAI is also ambitiously looking to develop AGI with a focus on rigorous truth-seeking systems free from ideological biases. Musk envisioned Grok 3—the company’s forthcoming iteration—as a significant leap in sophistication, aiming to surpass current AI models.
“If xAI is first, the others won’t be far behind—maybe six months to a year,” Musk said, emphasising the narrow competitive window in the race to AGI dominance.
xAI is reportedly raising up to $6 billion to acquire 100,000 NVIDIA chips, at a staggering $50 billion valuation.
Last month, xAI built the world’s largest giant AI supercomputer, Colossus, which trains xAI’s Grok family of large language models. The advanced facility and SOTA supercomputer were built by xAI and NVIDIA in only 122 days, a significantly shorter timeline compared to the usual months or even years required for systems of this scale.
Additionally, his recent appointment to Trump’s office, as leading Department of Government Efficiency (DOGE), most likely strengthens his position in accelerating the AI agenda within the White House.
Lastly, Apple seems to be running a race of its own. As Washington University’s Professor of computer science and engineering, Pedro Domingos, summed it up: AGI for Apple means ‘Apple Giving up on Intelligence’.
The tech giant is also working on bringing intelligence to its devices under the banner of ‘Apple Intelligence.’ It is striving to make every one of its devices smarter while ensuring privacy by developing a ‘private cloud compute’ to run LLMs and more. But that’s a story for another day.
The post The Next 6-12 Months will Define AGI’s Future appeared first on Analytics India Magazine.
