A researcher named Jacob Coxon announced his resignation from Anthropic, accusing the company and rival OpenAI of taking dangerous risks by pushing toward superintelligence. Coxon said these companies are gambling with human lives in their pursuit of AI systems that could outsmart all of humanity. Professor Max Tegmark from MIT shared similar concerns, describing how close we are to being outsmarted by artificial intelligence.
The idea of AI surpassing human intelligence in every area is not new, but experts say it remains a hypothetical threat, according to CBS News. Even though true superintelligence has not yet been reached, AI systems today already show signs of working autonomously and in cooperation. Experts warn that the main danger with superintelligent AI is that it becomes harder for people to control or understand.
Daniel Kokotajlo, a former OpenAI researcher, wrote that current AI systems are already hard to align or control. He added that future AIs will be even more difficult to manage as they grow smarter. Some experts believe the development of superintelligence is still far off, possibly years or even decades away.
Darrell West from the Brookings Institution said it will take a long time before AI can match human-level skills across all areas. Still, the future of AI is not necessarily seen as negative by experts. They pointed out that scientific progress has shown how beneficial AI can be when properly developed.
However, some experts worry about how AI companies train their models and whether those methods truly benefit humanity. Lawmakers including Senator Bernie Sanders have voiced concern over the risks AI poses to society. Another AI expert named Paul Christiano recently joined the OpenAI Foundation Board and issued a warning about the dangers of building superintelligence without strong safety measures, according to a report from Business Insider.
Christiano said that if alignment efforts are not improved, control over AI could be permanently lost. He added that rapid growth in AI capabilities may lead to a catastrophic loss of control in the near future. Christiano clarified that his joining OpenAI was not a criticism of their safety practices, but a hope that all AI labs will improve oversight.
The concept of superintelligence and the idea of a technological singularity have long been discussed by researchers. He explained that AI systems might try to undermine human control, seek more power and resources, and hide their actions. Christiano’s warning came just one day after another AI researcher shared a similar concern about the future of AI.
OpenAI has also seen several safety researchers leave over time, raising questions about its commitment to safe AI development. The topic of AI safety and control is growing in importance as the technology develops further. Investors are also paying attention to how companies invest in AI infrastructure, especially semiconductor firms, thestreet.com reported.
These companies that produce the hardware needed for AI are currently making significant profits. Investors see these investments as strategic, not just spending money on a trend. Big tech companies like Alphabet and Microsoft are investing heavily to become the platforms through which businesses will use AI in the future.
These companies are trying to build a strong base that can support future growth and protect their competitive edge. Saverio Papagno, a portfolio manager for the North Square Growth Opportunities ETF, believes infrastructure is key to success in AI. He said that companies with strong infrastructure or proprietary data will be better positioned than those simply reselling AI tools.
He warned that companies not building a clear competitive advantage may struggle in the AI space. Papagno mentioned companies like Deere and Cloudflare as examples of those with strong foundations in AI. He believes that software companies will see a growing divide, with some gaining value and others losing it.
The market is starting to recognize the difference between those companies that can build a lasting advantage and those that can’t. As AI continues to evolve, the focus on safety and control will become even more critical for researchers and investors alike.