In this week's artificial intelligence (AI) news, experts weigh in on Elon Musk's prediction that AI will soon surpass humans, companies are developing custom AI chips to cut costs, and analysts are AI-powered apps are booming in the mental health space, with the demand for high-quality data to power AI programs soon exceeding supply.
Elon Musk predicts AI will surpass human intelligence by 2026
Will AI outwit us? Elon Musk recently said that artificial intelligence (AI) could surpass the intelligence of even the brightest humans by as early as next year or 2026, sparking a heated debate among academics, technologists and ethicists. Ta.
In an interview with X, Tesla's CEO spoke about the accelerating pace of AI development aimed at achieving and exceeding human cognitive abilities. Experts are now weighing the possibility of Mr. Musk's prediction and the critical questions it raises about the nature of intelligence, ethical limits, and the future dynamics between humans and machines.
“Elon is right,” Yigit Yihraml, an AI researcher and founder of the AI-focused investment firm Vela Partners, said in a conversation with PYMNTS. “AI has already exceeded human intelligence in certain areas and will exceed human intelligence in many more areas, but not all.”
Why are so many AI companies developing their own chips?
AI chips are all the rage these days. Leading tech companies are actively developing their own custom chips to increase the efficiency and reduce costs of artificial intelligence (AI) operations.
Meta recently introduced an updated custom chip designed to enhance AI capabilities and reduce dependence on third-party suppliers such as Nvidia. The move is in line with Intel's recent announcement of advanced AI “accelerators” and reflects a similar effort by Google to produce AI chips in-house. Experts believe these chips could potentially power commercial AI applications.
“Custom chips lower the threshold for companies to train AI models tailored to specific customers and tasks, especially in specialized and advanced security scenarios, rather than simply using APIs from large language model providers. We’re going beyond that,” said Amrit Jasal, co-founder and CTO. His PYMNTS conversation with Egnyte, a company that develops AI-powered software for businesses.
Additionally, custom chips can significantly reduce AI-related costs for companies. Incorporating generative AI into business operations currently costs hundreds of thousands of dollars per month for bespoke solutions using fine-tuned open source models, according to software developer Itrex. . Last year, Nvidia CEO Jensen Huang highlighted the potential cost savings of custom AI chips.
AI may be hit by a data exhaustion spell
Data is the lifeblood of AI, but it's running out. Industry analysts have warned that growing demand for high-quality data, essential to powering AI conversation tools such as OpenAI's ChatGPT, could soon outstrip supply and hinder AI progress.
Increasing reliance on extensive datasets poses challenges to AI development. Such data is essential to improving models like ChatGPT, but the impending lack of data is causing concern among technical experts.
Training data for AI is scarce because it requires large amounts of diverse, high-quality, and accurately labeled data that reflects the real-world situations that models face. Collecting this data is often labor-intensive, requiring manual annotation by experts, gathering information from various locations, and careful attention to maintaining quality and eliminating bias.
Additionally, AI companies encounter significant copyright hurdles when collecting training data, and must carefully comply with legal requirements, obtain permissions, and filter content.
Jignesh Patel, a computer science professor at Carnegie Mellon University and co-founder of DataChat, a generative AI analytics platform, said, “Humanity is depleting this resource as quickly as LLM companies are depleting it. It is not possible to replenish.” “However, specialized LLMs rely less on publicly available data. For example, an LLM designed to automate the financial risk review process may In many cases, there is little or no public documentation available.”
The urgency to protect training data is underscored by lawsuits by authors and publishers against AI companies for illegally using their content to develop AI technology.
Making mental health more accessible with AI apps
The next therapist could be a chatbot. In the mental health field, interest in AI-powered apps is surging, with the promise of faster and more accessible care in an area of growing demand.
These AI applications have sparked debate among experts about their potential to fill gaps in care and concerns about their effectiveness and ethical implications. A key challenge is to determine whether AI can truly enhance the human element in mental health treatment.
Derek du Chene, CEO of Better U and developer of the mental wellness app, said PYMNTS AI could play a role in personalizing care. “AI is invaluable in mental health apps because it can provide support and interventions tailored to each user based on data, improving engagement and outcomes,” he said. “AI algorithms can detect patterns in behavior and mood over time, allowing for early intervention and preventive care.”
Du Chesne also said that AI provides round-the-clock support, which is essential in the absence of human therapists.
This trend is part of a broader movement towards the use of AI in customer service. PYMNTS reports that consumers are frequently using AI technology for a variety of applications, from preventing credit card fraud to processing returns with chatbots. Toronto-based company Meeranda is preparing to launch a visual AI designed to go beyond the capabilities of typical chatbots and emulate real-time human interaction, and not directly compete with platforms like ChatGPT. I am.
