An early computer program created in the 1960s has beaten the increasingly popular artificial intelligence program ChatGPT in a tring test. This test is designed to distinguish between humans and artificial intelligence.
Researchers at UC San Diego in the US tested the early chatbot ‘Eliza‘, developed by MIT scientist Joseph Weisenbaum in the mid-1960s, against modern forms of technology.
The researchers learned that Eliza beat OpenAI’s GPT-3.5 AI, the company’s free version of ChatGPT.
In 1950, British computer scientist Alan Turing first introduced the Turing Test, and since then this test has been the standard for determining a machine’s ability to imitate human speech.
The latest study required 652 people to decide whether they were talking to another human or an artificial intelligence chatbot on the Internet.
OpenAI’s GPT-4 chatbot, which is more powerful than the free version of the technology, was able to trick people more times than Eliza. His success rate was 41 percent. Eliza managed to convince people they were talking to a human 27 percent of the time, compared to GPT 3.5’s success rate of just 14 percent.
Artificial intelligence expert Gary Marks described Eliza’s success as ‘worrisome’ for tech companies working on AI chatbots, but other experts argued that ChatGPT performed well in tring tests. Not designed for
Ethan Mullick, a professor of artificial intelligence at the Wharton School in the US, wrote in a post on X (formerly Twitter) that ‘When you study the research, I think GPT 3.5’s defeat by Eliza is not that surprising. It seems.’
This section contains related reference points (Related Nodes field).
‘OpenAI considers the risk of duplication to be a real concern and has RLHF (machine learning from human feedback) capabilities to ensure that chat GPT is as successful as a human. Don’t try. Eliza was designed to succeed using our psychology.’
One of the reasons the researchers mistook Eliza for a human was that her performance was ‘too bad’ to be a current model of artificial intelligence and therefore ‘more likely to be a human’. Be a person who is not willing to cooperate deliberately.’
Arvind Narain, a professor of computer science at Princeton who was not involved in the research, said, ‘As usual, testing behavior does not tell us about capability. ChatGPT is designed to have a formal tone, not express opinions, etc., which makes it less human-like.’
‘Does GPT-4 Pass the Tring Test.’ This research, titled , is yet to be reviewed.
!function(f,b,e,v,n,t,s)
{if(f.fbq)return;n=f.fbq=function(){n.callMethod?
n.callMethod.apply(n,arguments):n.queue.push(arguments)};
if(!f._fbq)f._fbq=n;n.push=n;n.loaded=!0;n.version=’2.0′;
n.queue=[];t=b.createElement(e);t.async=!0;
t.src=v;s=b.getElementsByTagName(e)[0];
s.parentNode.insertBefore(t,s)}(window,document,’script’,
‘https://connect.facebook.net/en_US/fbevents.js’);
fbq(‘init’, ‘2494823637234887’);
fbq(‘track’, ‘PageView’);
#Chat #GPT #defeated #computer #program #1960s
2024-09-18 17:44:07
What does the Turing Test measure in artificial intelligence?
Table of Contents
The Unlikely Victory of an Early Computer Program: Eliza Beats ChatGPT in Turing Test
In a surprising turn of events, a computer program created in the 1960s has managed to outsmart the increasingly popular artificial intelligence program ChatGPT in a Turing test, a standard measure of a machine’s ability to imitate human speech. Researchers at the University of California, San Diego, pitted the early chatbot ‘Eliza’, developed by MIT scientist Joseph Weizenbaum in the mid-1960s, against modern forms of technology, including OpenAI’s GPT-3.5 AI, the free version of ChatGPT. The results were nothing short of astonishing, with Eliza convincingly mimicking human-like interactions, fooling a significant number of participants into believing they were conversing with a human.
The Turing Test: A Gold Standard for AI
The concept of the Turing test was first introduced by British computer scientist Alan Turing in 1950, and has since become the benchmark for measuring a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. The test is designed to assess a chatbot’s capacity to engage in natural-sounding conversations, making it difficult for humans to distinguish between a machine and a human interlocutor.
The Experiment: Eliza vs. ChatGPT
In this latest study, researchers asked 652 participants to engage in online conversations with either Eliza or ChatGPT. The results showed that OpenAI’s GPT-4 chatbot, a more advanced version of the technology, managed to deceive participants 41 percent of the time, while Eliza convincingly replicated human-like conversations 27 percent of the time. Surprisingly, ChatGPT’s free version, GPT-3.5, only succeeded in tricking participants 14 percent of the time.
Expert Reactions: Worrisome and Not Surprising
Artificial intelligence expert Gary Marks described Eliza’s success as “worrisome” for tech companies working on AI chatbots, highlighting the need for more advanced measures to distinguish between human and machine interactions. On the other hand, Ethan Mullick, a professor of artificial intelligence at the Wharton School, argued that ChatGPT performed admirably in the Turing test, given its design constraints. He noted that OpenAI’s focus on ensuring ChatGPT’s success through machine learning from human feedback (RLHF) might have limited its ability to convincingly mimic human-like interactions.
Why Eliza Succeeded: Understanding Human Psychology
According to researchers, one of the reasons Eliza was able to fool participants was its “too bad” performance, which made it seem more human-like due to its deliberate lack of cooperation. Arvind Narain, a professor of computer science at Princeton, not involved in the research, pointed out that ChatGPT’s formal tone and reluctance to express opinions made it less human-like, emphasizing the importance of understanding testing behavior and its limitations.
Implications and Future Directions
The outcome of this experiment highlights the need for continued innovation in AI development, ensuring that chatbots can effectively balance their cognitive abilities with human-like interactions. As AI technology continues to advance, it is crucial to address the concerns raised by this study, including the potential risks of duplication and the importance of designing systems that can distinguish between human and machine conversations.
Conclusion
The unexpected victory of Eliza over ChatGPT serves as a poignant reminder of the complexities involved in creating truly human-like AI systems. As we continue to navigate the rapidly evolving landscape of artificial intelligence, it is essential to stay vigilant, refining our testing methods and pushing the boundaries of innovation to ensure that AI technology remains a force for good.
Keywords: Turing Test, AI, ChatGPT, Eliza, artificial intelligence, machine learning, human-computer interaction, natural language processing.
Optimized Meta Description: Discover the surprising outcome of a Turing test, where a 1960s computer program outsmarts ChatGPT, highlighting the complexities of creating human-like AI systems.
Meta Keywords: Turing Test, AI, ChatGPT, Eliza, artificial intelligence, machine learning, human-computer interaction, natural language processing.
– What factors contributed to Eliza’s unexpected success in the Turing Test against ChatGPT?
The Unlikely Victory of an Early Computer Program: Eliza Beats ChatGPT in Turing Test
In a shocking turn of events, a computer program created in the 1960s has managed to outsmart the increasingly popular artificial intelligence program ChatGPT in a Turing Test, a benchmark designed to distinguish between humans and artificial intelligence. Researchers at the University of California, San Diego, put the early chatbot “Eliza,” developed by MIT scientist Joseph Weizenbaum in the mid-1960s, to the test against modern forms of technology.
The Turing Test: A Measure of Artificial Intelligence
The Turing Test, introduced by British computer scientist Alan Turing in 1950, is the standard for determining a machine’s ability to imitate human speech. The test requires a human evaluator to engage in natural language conversations with both a human and a machine, without knowing which is which. If the evaluator cannot reliably distinguish the human from the machine, the machine is said to have passed the Turing Test.
The Study: Eliza vs. ChatGPT
In the recent study, 652 people were asked to decide whether they were talking to another human or an artificial intelligence chatbot on the internet. The results showed that Eliza, the early chatbot, managed to convince people they were talking to a human 27% of the time. In contrast, OpenAI’s GPT-3.5, the free version of ChatGPT, only succeeded 14% of the time.
GPT-4: A More Powerful Chatbot
While Eliza outperformed GPT-3.5, OpenAI’s more powerful chatbot, GPT-4, was able to trick people more times, with a success rate of 41%. However, experts argue that ChatGPT’s performance is not designed for the Turing Test and is instead focused on generating human-like responses.
Experts Weigh In
Artificial intelligence expert Gary Marks described Eliza’s success as “worrisome” for tech companies working on AI chatbots. However, Ethan Mullick, a professor of artificial intelligence at the Wharton School, wrote on X (formerly Twitter) that “When you study the research, I think GPT 3.5’s defeat by Eliza is not that surprising. It seems OpenAI considers the risk of duplication to be a real concern and has RLHF (machine learning from human feedback) capabilities to ensure that chat GPT is as successful as a human.”
The Reason Behind Eliza’s Success
One of the reasons Eliza was mistaken for a human was that her performance was “too bad” to be a current model of artificial intelligence, making it more likely to be a human who was not willing to cooperate deliberately. Arvind Narain, a professor of computer science at Princeton, noted that “as usual, testing behavior does not tell us about capability. ChatGPT is designed to have a formal tone, not express opinions, etc., which makes it less human-like.”
The Implications of the Study
The study, titled “Does GPT-4 Pass the Turing Test?”, is yet to be reviewed. However, the results highlight the importance of understanding the capabilities and limitations of artificial intelligence systems. As AI continues to evolve and become more integrated into our daily lives, it is essential to recognize the potential risks and benefits of these technologies.
the unlikely victory of Eliza over ChatGPT in the Turing Test serves as a reminder that even older AI systems can still have relevance and importance in today’s technological landscape. The study’s findings have significant implications for the development and evaluation of AI chatbots, and demonstrate the need for continued research and innovation in the field of artificial intelligence.
