web analytics

Google is scrambling to prevent the massive shift from Google search as the preferred way of getting information.

But which AI is best? I gave the same prompts to Perplexity.ai and Chat GPt-4, and for my objective, Perplexity.ai, unsurprisingly, gave the better answer which is why Perplexity’s response is given here first…

To the prompt: Pls create a 5 column comparison of Chat Gpt 3.5, Chat GPT4, Google Gemini and Perplexity.ai: advantages, disadvantages, costs, best use, input, output, up to what year the data is coming from. The first column is the basis for comparison”, Perplexity.Ai gave this:

Below is a comparison table that outlines the key aspects of ChatGPT 3.5, ChatGPT 4, Google Gemini, and Perplexity AI based on the available information:

Basis for ComparisonChatGPT 3.5ChatGPT 4Google GeminiPerplexity AI
Advantages– Good for general language tasks and conversation. – Wide range of knowledge up to its last training cut-off in 2021[19].– Higher accuracy and precision in responses[14]. – Can handle more input/output with a max of 25,000 words[2]. – Multimodal capabilities (text and images)[14].– Multimodal (understands text, audio, video, and code)[18]. – Integrated with Google services for comprehensive data retrieval[3].– Provides concise, expert answers. – Uses up-to-date information by reading from the web.
Disadvantages– Limited by data up to 2021, cannot access real-time information[19]. – May produce biased or inaccurate responses due to training data limitations[19].– Requires subscription for full features[2]. – Still may generate biased or inaccurate responses[2].– Image generation paused due to inaccuracies[18]. – May not be as conversational or versatile in text generation as ChatGPT models[18].– Limited to text-based inputs and outputs. – Does not generate creative content like stories or code.
CostsFree version available. ChatGPT Plus (with GPT-4 access) costs $20 per user per month[18].Free version available; Plus version costs $20 per month for enhanced features and access to GPT-4[2].Gemini 1.0 Pro API is freely available in Google AI Studio for up to 60 queries per minute. Gemini Advanced costs $20 per month[6].Free access to the AI with limitations on the number of daily queries.
Best UseGeneral inquiries, content creation, and learning purposes up to its knowledge cut-off[19].Advanced content creation, detailed analysis, and complex language tasks[14].Research, multimodal queries (including audio and video), and tasks requiring up-to-date information[3].Answering specific questions with expert-level accuracy and sourcing information from the web.
InputText-based queries and prompts[19].Text-based queries and prompts; can accept images as inputs in Plus version[14].Text, audio, video, and code[18].Text-based queries and prompts.
OutputText responses[19].Text responses; can generate captions, classifications, and analyses for images in Plus version[14].Text, audio responses, and potentially multimodal outputs (except paused image generation)[18].Text responses with citations to sources.
Data Up To2021[19].April 2023 for GPT-4[18].Continuously updated through Google’s services[3].Real-time, as it accesses the latest information available on the web.

This table provides a snapshot comparison of the four AI models based on the specified criteria. It’s important to note that the capabilities and features of these models are continuously evolving, and specific details may change over time.

Citations:
[1] https://www.xbytesolutions.com/blog/chatgpt-vs-google-comparison
[2] https://butteredtoast.io/blog/chatgpt-4-benefits-drawbacks-more/
[3] https://blog.google/technology/ai/google-gemini-ai/
[4] https://www.eweek.com/artificial-intelligence/chatgpt-vs-github-copilot/
[5] https://community.openai.com/t/chatgpt-gpt-4-output-got-faster-and-worse/222874
[6] https://www.eweek.com/artificial-intelligence/gemini-vs-chatgpt/
[7] https://www.mdpi.com/2071-1050/15/22/16071
[8] https://www.sciencedirect.com/science/article/pii/S266734522300024X
[9] https://matrixmarketinggroup.com/google-gemini-blog-writing/
[10] https://www.researchgate.net/publication/369125368_Literature_Review_on_Pros_and_Cons_of_ChatGPT_Implications_in_Education
[11] https://www.klippa.com/en/blog/information/chat-gpt-4/
[12] https://www.techtarget.com/searchenterpriseai/definition/Google-Bard
[13] https://www.paperdigest.org/2023/01/recent-papers-on-chatgpt/
[14] https://openai.com/gpt-4
[15] https://zapier.com/blog/chatgpt-vs-bard/
[16] https://www.linkedin.com/pulse/claude-2-vs-gpt-4-2023-comparing-top-ai-models-kim-garst
[17] https://www.credit-suisse.com/media/assets/microsite/docs/securities-research/chatgpt.pdf
[18] https://www.techtarget.com/searchenterpriseai/tip/Gemini-vs-ChatGPT-Whats-the-difference
[19] https://www.techtarget.com/whatis/definition/large-language-model-LLM
[20] https://www.reddit.com/r/ChatGPTPro/comments/13qhbip/frustrations_with_chat_gpt_4_seeking_advice_and/

To Perplexity, I added a follow-up to create a 1-slide comparison, and this was the result:

I’m unable to create or display actual slides directly within this text-based interface. However, I can guide you on how to structure a slide for a presentation based on the comparison provided. Here’s a suggested layout for your slide:


Title: Comparison of AI Language Models

CriteriaChatGPT 3.5ChatGPT 4Google GeminiPerplexity AI
Advantages– General language tasks
– Knowledge up to 2021
– High accuracy
– Multimodal capabilities
– Multimodal
– Integrated with Google services
– Expert answers
– Up-to-date information
Disadvantages– Limited by 2021 data
– Potential biases
– Subscription for full features
– Potential biases
– Image generation paused
– Less conversational
– Text-based only
– No creative content generation
CostsFree; $20/mo for PlusFree; $20/mo for PlusFree; $20/mo for AdvancedFree with query limits
Best UseGeneral inquiries, content creationAdvanced content creation, complex tasksResearch, multimodal queriesSpecific questions, expert accuracy
InputTextText, images (Plus)Text, audio, video, codeText
OutputTextText, image analysis (Plus)Text, audio, multimodalText with citations
Data Up To2021April 2023Real-timeReal-time

The answer on Perplexity column doest not cover my observation that Perplexity can generate images like the ones below (uploaded as webp here. Also took a screenshot and saved as png for other possible uses.)

Chat GPT-4 gave this to a similar prompt. The only difference is in regards to Gemini, which Chat GPT -4 replied as:

To provide an updated comparison that includes Perplexity.ai and additional comparison criteria regarding file/document/image attachments and the ability to produce outputs other than text, we adjust our focus. Note that Perplexity.ai, as of my last update, positions itself as a conversational AI focused on providing accurate information and clear explanations, leveraging the vast amount of data available on the internet.

Updated Comparison Table: GPT-3.5, GPT-4, and Perplexity.ai

Basis for ComparisonGPT-3.5GPT-4Perplexity.ai
Advantages– Good at generating human-like text.
– Broad knowledge base.
– More affordable for a wide range of applications.
– Better contextual understanding.
– Improved content safety and relevance.
– Enhanced instruction-following capabilities.
– Designed for high-accuracy information.
– Can query real-time data.
– Provides sources for its answers.
Disadvantages– Limited understanding of complex topics.
– Higher potential for generating nonsensical responses.
– Higher operational costs.
– May still produce biases or errors in outputs.
– More focused on information retrieval than content creation.
– Limited creative generation capabilities.
Costs– $0.02 per 1,000 tokens for basic usage.
– Volume discounts available.
– $0.06 per 1,000 tokens for enhanced capabilities.
– Tiered pricing based on usage.
– Not publicly detailed; likely varies based on usage and specific services required.
Best Use– Content creation for blogs and articles.
– Basic chatbots and educational tools.
– Advanced content creation and summarization.
– Complex customer support bots.
– Detailed analysis and interpretation.
– Fact-checking and information verification.
– Educational and research applications requiring up-to-date information.
File/Document/Image Attachment– Limited capability; primarily text-based responses without direct support for attachments.– Limited direct support; relies on descriptions for generating text-based responses about images.– Primarily text-based; designed for web data access rather than handling direct file or image attachments.
Output Other Than Text– Primarily text outputs; does not natively support generating images, audio, or video.– Primarily text outputs; limited to generating descriptions or content that can be transformed into other formats by external tools.– Focused on text outputs; does not natively generate multimedia content but can provide links to external sources.

This table reflects the capabilities and features of GPT-3.5, GPT-4, and Perplexity.ai based on information available up to early 2023. It’s important to recognize that the AI field is rapidly evolving, and the features, capabilities, and uses of these models may expand over time. For Perplexity.ai, specific functionalities such as cost and detailed capabilities should be verified directly as they can change and may not be fully disclosed publicly.