Anthropic AI, founded by former OpenAI researchers, has unveiled its latest and advanced chatbot, Claude 2, taking direct aim at its competitors like ChatGPT and Google Bard. With only five months passing since the launch of its predecessor, Claude 2 comes with several remarkable improvements, offering longer and more nuanced responses, and displaying superior performance, impressively scoring in GRE reading and writing exams.
One of Claude 2’s outstanding features is its ability to digest an impressive 100,000 tokens, equivalent to approximately 75,000 words, in a single prompt. This substantial leap from Claude’s previous 9,000 token limit provides the AI with a unique advantage, enabling it to deliver responses in a highly contextual and improved manner.
Claude 2 has undergone rigorous testing in various domains, including law, mathematics, and coding. According to Anthropic, the chatbot achieved an impressive 76.5% in the multiple-choice section of the Bar exam, surpassing GPT-3.5’s score of 50.3%. Additionally, Claude 2 outperformed over 90% of graduate school applicants in the GRE reading and writing exams. Furthermore, it demonstrated its advanced computational abilities by scoring 71.2% on the Codex HumanEval Python coding test and 88.0% on GSM8k grade-school math problems.
One of Claude’s distinctive traits is its “constitution,” which draws inspiration from the Universal Declaration of Human Rights. This constitution allows the AI to self-improve without human feedback, identify and rectify improper behavior, and adapt its conduct accordingly, demonstrating an ethical approach to AI development.
In comparing Claude 2 to its competitors, ChatGPT and Google Bard, let’s examine some key specifications:
ChatGPT vs Bard vs Claude
In the battle of AI prompts, we pitted three powerful contenders against each other: ChatGPT, Bard, and Claude. Here’s a rundown of how they fared in different scenarios:
Understanding foreign languages:
We tested their language comprehension by asking the meaning of a Spanish slang phrase. Claude proved to be the most accurate, providing a careful explanation. ChatGPT offered a satisfactory response, while Bard declined to answer, citing its inability to speak Spanish. However, when we rephrased the prompt, Bard delivered a better response than ChatGPT, although not as extensive as Claude’s.
Up-to-date information:
To assess their real-time information capabilities, we inquired about the price of Bitcoin on that day. Both ChatGPT and Claude lacked internet connectivity, rendering them unable to provide up-to-date information. However, Claude hallucinated a response with incorrect details, potentially misleading users. Google Bard, on the other hand, furnished the correct and current information, showcasing its up-to-date data retrieval abilities.
Context handling:
In this test, we challenged the models with a large text excerpt from the Bible and asked specific questions from the provided text. As expected, only Claude excelled, managing to process the extensive prompt and offering an accurate reply. Despite taking around 2 minutes to analyze the content, Claude demonstrated its adeptness at handling context-rich tasks, without resorting to cheating or skirting the question.
Non-verbal abilities:
While AI language models are not primarily designed for math tasks, we decided to put them to the test. We asked the models to create a payment plan for clearing credit card debts and rank which cards to use and avoid. Claude delivered the most comprehensive plan. However, it made a mistake by recommending prioritizing the card with the highest APR, potentially leading to suboptimal financial decisions.
Strengths and weaknesses
Claude 2:
Strengths:
- Impressive Context Handling: Claude 2shines with its ability to handle extensive contexts up to 100,000 tokens, setting it apart in text-intensive tasks.
- Versatile Performance: It demonstrates superiority in diverse fields like law, mathematics, and coding, boasting exceptional scores in standardized tests, showcasing its versatility.
- Autonomous Improvement: Claude 2 can self-improve and adapt without human feedback, displaying a remarkable ability to evolve on its own.
- Privacy and VPN Support: With VPN browsing and the option to delete conversations, Claude 2 ensures user privacy and security.
- Slack Integration and API Support: The chatbot’s integration with Slack and extensive API support makes it a valuable addition to productivity tools.
Weaknesses:
- Limited Availability: One of its drawbacks is its temporary availability restricted to the US and UK, potentially leaving users in other regions without access.
- No Internet Connection: Claude 2’s lack of internet connectivity means it cannot provide real-time data, making it less suitable for up-to-date information queries.
- Potential for Incorrect Information: In certain situations, Claude 2 might provide incorrect answers in complex tasks, even though it convincingly presents them.
ChatGPT:
Strengths:
- Wide Availability and Language Support: ChatGPT is the most widely accessible among the three models, supporting over 80 languages, making it user-friendly for a global audience.
- API Support and Plugin Store: The ChatGPT Plus version offers additional features, including API support and access to a plugin store, enhancing its capabilities.
Weaknesses:
- Context Limitations: Compared to Claude 2, ChatGPT has limited context handling capabilities, restricting its effectiveness in certain text-heavy scenarios.
- Quality Difference in Free Version: The free version lacks the added features and performance of the paid version, making it a somewhat limited option.
- Paused Web Browsing: ChatGPT’s web browsing feature is temporarily paused, making it unsuitable for real-time data queries.
- Potential for Inappropriate Outputs: In some complex tasks, ChatGPT may generate inappropriate results, which can be problematic in specific use cases.
Google’s Bard:
Strengths:
- VPN and Real-Time Data Support: Bard supports VPN browsing and offers real-time data access, making it suitable for up-to-date information queries.
- Integration Plans: The intention to integrate with Google Suite and offer a plugin store opens possibilities for seamless user experiences.
Weaknesses:
- Limited Language Support: Compared to ChatGPT, Bard supports fewer languages, which could be a drawback for international users.
- API Limitations: Bard’s API access is restricted, potentially hindering its integration with third-party applications.
- Context Handling Challenges: While capable, Bard’s context handling capabilities are less impressive than Claude 2’s, impacting its performance in certain tasks.
- Generic Responses in Complex Tasks: In some complex tasks, Bard’s responses might be generic and unhelpful, requiring a trade-off to reduce the risk of hallucinations.
Conclusion
With the ever-expanding array of options in the field of AI language models and chatbots, there’s no need to limit oneself to being a devoted ChatGPT enthusiast or exclusively align with Google’s offerings.
If the $20 price tag for ChatGPT Plus gives you pause, consider exploring Claude as a viable alternative. Claude’s capabilities are on par with GPT-4, and it’s likely to outperform GPT-3.5, which is available in the free version of ChatGPT. For most users, Claude proves to be a better choice than Google Bard.
A standout feature of Claude is its ability to analyze PDFs and files with various extensions, a functionality that’s comparable to the paid plugins available in the ChatGPT Plus subscription. So, before committing to ChatGPT 4 and investing financially, it’s worth giving Claude a try, as it might save you some money while offering comparable functionality.
Each AI chatbot has its own strengths and weaknesses, making them more suitable for specific tasks. Claude excels at handling large datasets but may not be ideal for tasks requiring real-time data access. On the other hand, ChatGPT boosts creativity, making it an excellent option for tasks demanding specific language support. Its plugin store, though requiring payment, is highly valuable for those willing to invest. Bard, meanwhile, stands out for its factual accuracy and real-time data capabilities, making it a reliable choice for certain use cases.
In the end, there’s no need to limit yourself to just one option. Embrace the diversity of AI chatbots available and utilize them all based on your specific needs and requirements. Each of them brings unique strengths to the table, empowering users to make the most of the rich landscape of AI language models and chatbot technology.