Artificial intelligence (AI) programs have been gaining momentum in recent years, with investments and expectations on the rise. However, there are new challenges and limitations to consider as these programs evolve. According to a prediction by IDC, enterprise spending on […]
Artificial intelligence (AI) programs have been gaining momentum in recent years, with investments and expectations on the rise. However, there are new challenges and limitations to consider as these programs evolve.
According to a prediction by IDC, enterprise spending on AI is expected to reach $143 billion by 2027, up from $16 billion this year. This surge in investment has contributed to the growth of tech companies like Nasdaq Composite, which has seen a 36% increase in value. However, there is a need for caution as the potential of AI technology may be overestimated.
OpenAI, a pioneering startup, aspires to achieve artificial general intelligence (AGI) that is comparable to human intelligence. However, it is important to note that AI models predict, rather than understand. This limitation raises questions about the ability of AI to achieve a level of general intelligence similar to that of humans.
Text generation using language models depends on the data used for training. While impressive results are achieved when models reflect recurring concepts, they struggle with new situations and tasks that are outside their “envelope.” This explains why Google DeepMind’s weather forecasting AI model outperforms existing models in certain cases. AI often faces difficulties in recognizing its own errors. Simply requesting a correction does not guarantee a more accurate response. In a study conducted by Originality.AI, every analyzed AI model showed errors, with OpenAI’s ChatGPT-4 providing inaccurate answers in nearly a third of cases.
Financial executives are pragmatically seeking ways to apply AI tools to achieve their goals. From analyzing employee performance to waste collection scheduling, the fields of application are diverse. However, results have varied. A study by the National Bureau of Economic Research on the use of AI chatbots in the workplace showed a 14% improvement in productivity. However, customer support agents achieved limited results, with only new hires and those with low levels of expertise making progress.
The limitations of AI will become more evident as new AGI tools are introduced in 2024. This will put pressure on suppliers to address the ultimate unresolved question: price. According to McKinsey, a renowned consulting company known for its exceptional forecasts, AI could add over $4 trillion to corporate profits. However, the precise determination of price remains unclear. Without this information, companies cannot accurately predict the financial benefits they can achieve through the use of AI, just as AI itself cannot accurately predict the weather.
The team at Lex wants to hear more opinions on this topic from readers. Please let us know in the comments below if you believe that AGI can reach superintelligence.