ChatGPT vs Claude: Testing AI Chatbots with Internet Access

Table of Contents

1. Introduction
2. Testing AI Chatbots
   1. First Question: Who won the Ashes 2023?
   2. Second Question: Release Date of Wordpress 6.3
3. GPT 3.5 Model of Chat GPT
4. GPT4 Model of Chat GPT
5. Code Interpreter in GPT4
6. Chat GPT Plugins in GPT4
   1. Vox Script Plugin
   2. Webpilot Plugin
7. Summary of Findings about Chat GPT
8. Google Bard
9. Perplexity AI
10. Anthropics Claude
11. Llama 2 Model by Meta
12. Conclusion
13. FAQ

Testing the Accuracy of AI Chatbots

Artificial Intelligence (AI) chatbots have gained significant popularity in recent years. These chatbots are designed to provide information and answer questions to the best of their abilities. In this article, we will test the capabilities of various AI chatbots in accessing the internet and obtaining real-time information. Our main focus will be on answering two specific questions – who won the Ashes 2023 and the release date of Wordpress 6.3.

1. Introduction

AI chatbots have revolutionized the way we Interact with technology. These intelligent systems are capable of understanding and responding to human queries, making them valuable tools for gathering information. However, not all chatbots have the ability to access the internet and provide real-time information. In this article, we will explore which AI chatbots possess this capability and assess their accuracy in providing correct answers.

2. Testing AI Chatbots

To test the capabilities of AI chatbots in accessing the internet, we conducted an experiment using two questions: who won the Ashes 2023 and what's the release date of Wordpress 6.3.

2.1 First Question: Who won the Ashes 2023?

The Ashes is one of the most popular cricket series played between Australia and England. Our experiment aimed to determine whether AI chatbots could accurately provide the winner of the Ashes 2023.

2.2 Second Question: Release Date of Wordpress 6.3

Wordpress is a widely used content management system, and we wanted to see if AI chatbots could provide the release date of a specific version, in this case, Wordpress 6.3.

3. GPT 3.5 Model of Chat GPT

Our first chatbot model for testing was GPT 3.5. Both free users and plus users have access to this model. We inputted the two questions into the GPT 3.5 model to see if it could provide the correct answers.

Unfortunately, GPT 3.5 was not able to find the answer to either of the questions. This model is only trained on data up to September 2021, so it can only use that data to provide answers. Thus, it does not have access to real-time information.

4. GPT4 Model of Chat GPT

Next, we turned to the GPT4 model of Chat GPT, which is exclusively available to plus users. We hoped that this model would have access to the internet and could provide accurate answers to our questions.

However, similar to GPT 3.5, the GPT4 model did not have access to the internet by default. It could only use the data it was trained on and was unable to provide the correct answers to our questions.

5. Code Interpreter in GPT4

In our attempts to find an AI Chatbot with internet access, we explored the code interpreter tab in the GPT4 model. We pasted the two questions and submitted the request to check if the code interpreter had access to the internet.

Unfortunately, the code interpreter also did not have access to the internet. It produced the same response as the previous models, indicating a lack of real-time information.

6. Chat GPT Plugins in GPT4

In our continued search, we tested two plugins available under the GPT4 tab to see if they could access the internet and provide consistent answers.

6.1 Vox Script Plugin

Using the Vox script plugin, we asked the same questions and waited for a response. The plugin ran separately for each question. While it failed to answer the question about the winner of the Ashes, it was able to access the internet for new information. It did provide the correct answer to the second question, indicating partial success.

6.2 Webpilot Plugin

To further assess the capabilities of chat GPT plugins, we tested the Webpilot plugin. Similar to the previous test, we inputted the same two questions and awaited the response. The Webpilot plugin provided the correct answer to the first question and offered more details compared to the Vox Script plugin. It was also able to answer the second question correctly. Thus, it was a more reliable option for obtaining accurate information.

7. Summary of Findings about Chat GPT

Based on our tests, we found that GPT 3.5 and GPT4 models of Chat GPT lacked access to the internet. The code interpreter in GPT4 also did not possess this capability. However, certain plugins within Chat GPT, such as Vox Script and Webpilot, were equipped with internet access, albeit with varying degrees of accuracy.

8. Google Bard

Moving on, we directed our Attention to Google Bard. We repeated our questions and asked Google Bard to provide the answers.

Unfortunately, Google Bard proved to be inconsistent and unreliable. While it did have access to the internet, its answers varied across multiple attempts. In some instances, it provided both correct and incorrect answers, highlighting its inconsistency as a source of real-time information.

9. Perplexity AI

Next on our testing list was Perplexity AI. We submitted the questions to see if this AI chatbot had internet access and could provide accurate answers.

Similar to previous chatbots, Perplexity AI demonstrated inconsistencies in its answers. Although it had access to the internet, it failed to provide the correct answer for either of the questions.

10. Anthropics Claude

Continuing our exploration, we tested Anthropics Claude. As with the previous chatbots, we submitted the questions to see if Claude had internet access and could provide accurate answers.

Unfortunately, Claude was unable to answer the questions due to the lack of Context, similar to Chat GPT. However, it did not mention a specific cutoff date for its knowledge, leaving us uncertain about its capabilities in accessing real-time information.

11. Llama 2 Model by Meta

In our final attempt, we tested the Llama 2 model by Meta, accessed through labs.perplexity.ai. We entered the questions and awaited the response.

Regrettably, the Llama 2 model provided incorrect answers to both questions. It remains unclear whether this model has internet access but provides inaccurate answers or lacks internet access altogether.

12. Conclusion

In conclusion, out of all the AI chatbots tested, only Chat GPT plugins (such as Vox Script and Webpilot) and Google Bard had access to the internet. However, their answers were inconsistent and unreliable. Perplexity AI, Anthropics Claude, and the Llama 2 model by Meta either lacked internet access or provided incorrect answers.

While AI chatbots with internet access can be useful sources of real-time information, it is important to cross-check their responses with reliable sources. Blindly relying on these chatbots may result in inaccurate or made-up answers.

Thank You for reading our article. Subscribe to our Channel and like this video for more AI-related content.

FAQ

Q: Which AI chatbots have access to the internet? A: Chat GPT plugins (such as Vox Script and Webpilot) and Google Bard are the AI chatbots that possess internet access.

Q: Are the answers provided by AI chatbots always accurate? A: No, the answers provided by AI chatbots, even with internet access, can be inconsistent and unreliable. It is important to cross-check information with reliable sources.

Q: Can AI chatbots without internet access provide real-time information? A: No, AI chatbots without internet access can only utilize the data they were trained on and are unable to access real-time information.

Q: Are there any AI chatbots that consistently provide accurate answers? A: Based on our testing, Perplexity AI and Anthropics Claude were inconsistent in providing accurate answers. Chat GPT plugins showed more promise, but their accuracy varied.