Search engines already feel like second choice to GPT-4
Thu, 06 Apr 2023 by garethbrown
Having spent a week and a half with GPT-4, I've found my reflex when looking for information is to go to GPT-4 before Google (or any other traditional search engine). These are the main factors:
1. The contextual awareness for follow up questions in GPT-4 is great
Rather than having to tune individual search strings for a search engine that narrow the results according to the specifics of the problem at hand, I'm finding that what's needed for GPT-4 tends to be a single longer prompt at the outset of the 'search', but then improving the response requires only short prompts to add additional detail, much like a human conversation. This somehow feels better, like once you've started you have enlisted the help of someone who can help and everything from there on is a team effort.
2. Consolidation of information
In both English language and code sample responses, GPT-4 show's an ability to condense and organise information from disparate sources and present them in a coherent response. This is labour saving in that normally the full set of information would be gathered over multiple traditional search engine queries, which can be tiring for complex information gathering exercises.
3. No Ads (yet)
GPT-4 is only available to most people at the time of writing under the ChatGPT Plus subscription. This is a paid service and therefore there is no need for advertising support. This results in a less noisy interface with no distractions, and the feeling that there is no ulterior motive for providing you with a service. This will likely only remain a feature of paid services. The potential for context aware ads being displayed alongside chats with LLMs must be obvious to anyone working in online advertising.
4. No SEO spam
Even if traditional search engines respond with relevant and useful results, the pages you land on often have laborious introductions that unintentionally or intentionally pad the content with keywords that influenced the search indexing for the page. This takes extra mental effort to parse while you look for the information that you really need. Google search helps but highlighting search result text on the landing page, though this is inconsistent and is still not as good as not having the page bloat there in the first place.
And all the other reasons ...
As I'm often finding myself doing even when I have a solution to a problem in hand, I'll run it by ChatGPT as well. As often is the case, ChatGPT was able to confirm and add to my thoughts:
List advantages that large language models such as GPT have over traditional search engines.
Large language models like GPT offer several advantages over traditional search engines:
Natural language understanding: GPT models have a deep understanding of human language and can comprehend complex questions, allowing users to communicate with the system more naturally and effectively.
Context awareness: GPT models can maintain context in a conversation, understanding the relationship between multiple queries and providing coherent, contextually relevant responses.
Conversational ability: GPT models are designed to engage in multi-turn conversations, which enables more dynamic interactions compared to the one-way query input of traditional search engines.
Improved query interpretation: GPT models can often understand vague or poorly phrased queries, deciphering user intent and providing helpful answers despite the ambiguity.
Personalized responses: Large language models can adapt their responses based on user preferences, providing more personalized and relevant information.
Generating content: GPT models can generate text, which allows them to provide more detailed explanations, examples, or creative suggestions when responding to queries.
Multilingual capabilities: GPT models can work with multiple languages, enabling users from different linguistic backgrounds to access information and engage in conversations.
Summarization and synthesis: GPT models can process vast amounts of information and provide concise summaries or synthesize insights, which is particularly helpful for complex topics or when users need a high-level overview.
Adaptability: GPT models can be fine-tuned for specific tasks or domains, making them more versatile and capable of handling a wide range of applications.
Continuous learning: As large language models like GPT continue to be trained on new data, they can improve their knowledge base and performance over time, staying up-to-date with the latest trends and information.
The use of any information, code samples, or product recommendations on this Website is entirely at your own risk, and we shall not be held liable for any loss or damage, direct or indirect, arising from or in connection with the use of this Website or the information provided herein.