So if I train an agent with my company data and base it on any one of the LLM's, will that be publicly available data or can I somehow restrict it so that only my company folks get to see it as it will hav e information specific to my company? 🤔🤔
You can build agents with any model in theory, which means you could use some open source ones like Llama and use it locally on your computer. In short, meaning you dont give any large company your personal data. But for many of these companies (like OpenAI) they say they dont train in your data if you use their API or paid plans. Now, whether you believe that is true is another question but in theory it shouldnt be used by those companies, but it may be seen by them.
Thank you for your informative video comparing the three main AI assistants: Claude, ChatGPT, and Gemini. As a sales and market manager, I found your insights particularly valuable for my daily tasks, such as report writing, analysis, planning, and improving my English writing skills. After conducting my own tests using the free versions of Claude and ChatGPT, I have decided to subscribe to Claude, as you pointed out that ChatGPT's writing tends to have a more noticeable AI footprint, while Claude's output resembles human writing more closely. This is a crucial factor for me, as I want my reports and analyses to sound professional and human-like. However, I have noticed that Claude lacks internet searching capabilities, which has made me consider subscribing to ChatGPT-4 as well. The ability to access and incorporate current information from the internet could be highly beneficial for my work. As you mentioned, cost is always a consideration, and I am unsure if I should invest $40 for both Claude and ChatGPT subscriptions. Given my work requirements and the importance of human-like writing, I am still leaning towards selecting Claude for my professional needs. Thank you once again for sharing your experiences and helping me make a more informed decision. Your video has been a great resource, and I appreciate the time and effort you put into creating this comparison.
Alfie, Thank you for sharing specific examples, that demonstrate the advantages of using perplexity over other AI models in research. I especially found it helpful that you explained how most models pull from a static database of information, whereas perplexity pulls from a dynamic open ended database. Also your accent is sexy AF.
There's nothing stopping you from making them fully agentic now. They totally can. the problem is that the results at the moment are often subpar. Mainly because the current models capability to reason (i.e. know what steps to take) is inferior. Often because models use general knowledge to come up with reasoning which may not be accurate. For example, if the user asked an agent to book me a movie ticket, it could totally do that automatically by looking at your calendar, seeing when you're free, seeing what films are available at that time, and booking a movie. But the level of reasoning needed to know the right actions might be basic. The agent might say "To plan this movie I need to know when Alfie's free, and what films are available at that time", but the model may not know to reason with things like "Is it for a date with his wife, or a movie with his kids" or "Should I prioritize the cinema with the films he likes, or should i prioritize the cinema with the best popcorn?" The reasoning part is where they fall down today.
Wow, really Your video making is very good. Your content ideas awesome, if you Give me permission, I will share some information with you to increase the channel. You Get lots of subscribers and views
Except for the RAG function Hugginface models work pretty well and they are free. Also you can build agents based on your information but it has to be online because you can't upload documents yet. I tried the UFC question and this was the result: who won in the max holloway vs justin gaethje UFC fight? According to the search results, Max Holloway won the fight against Justin Gaethje via knockout (punch) in Round 5, with only 1 second remaining on the clock. what were the score cards like for the max holloway vs justin gaethje UFC fight? According to the text, the judges' scorecards for Max Holloway vs Justin Gaethje were: 37-39, 37-39, 38-38 These scores indicate that Max Holloway was leading throughout the fight, with a dominating fifth round likely only increasing his lead. However, one of the judges, Junichiro Kamijo, scored the fight as a draw, which received some backlash online. You need to check the veracity of the information and start new conversations for different subjects but it works quite well. It uses Llama 3 Large model or Mistral Large model wich are quite good. I'm sure Perplexity is more accurate but hugging face works quite well considering the price...
Creating project, would you like to be a judge? 1. CodeCraft Duel: Super Agent Showdown 2. Pixel Pioneers: Super Agent AI Clash 3. Digital Duel: LLM Super Agents Battle 4. Byte Battle Royale: Dueling LLM Agents 5. AI Code Clash: Super Agent Showdown 6. CodeCraft Combat: Super Agent Edition 7. Digital Duel: Super Agent AI Battle 8. Pixel Pioneers: LLM Super Agent Showdown 9. Byte Battle Royale: Super Agent AI Combat 10. AI Code Clash: Dueling Super Agents Edition
Hey there, I've been analyzing your RU-vid channel over the past three days, and it's clear that your RU-vid Channel and Video SEO are currently struggling to rank. Are you aware of this issue? If it persists, your audience may have difficulty finding you through RU-vid search, and it's crucial to aim for a place on the first page. If you're interested in optimizing your video and channel SEO, you'll need the right keywords to boost your podcast RU-vid channel's visibility. I specialize in keyword research and can provide comprehensive video and channel optimization services to improve your ranking. How does that sound?
In a nutshell, yes. BUT the big difference is it synthesizes information across those sources so its great for obscure questions that no single link can answer. For example, “who is the most watched basketball player with the most shoe sales”. Those two statistics could be across two different sources, but perplexity can give you the answer in one go. Its great for making detailed research quicker without having to crawl the internet manually.
While OpenAI doesn't provide any sources at all, search engines (like Bing, DuckDuckGo, Google etc.) provide only sources. Perplexity combines that, making your research a little easier. Additionally, you can ask Perplexity e.g. to verify the sources, or critique its own answer that way making it "better" (more complete). Additionally one can ask for a certain output format (e.g. MarkDown etc.) thus saving time.
Google the king of search and youtube and all the resources they have to hire the best of the best; i dont see why they do not adopt a perplexity type of search on their search engine for free like they do with their regular search. This would render all Ai search bots redundant.
@@AlfieMarsh They can keep both the regular search and AI. Just add on the google page Google search and AI search and customers can click which one they they want to use.. I would not get rid of the regular search as it serves its purpose.
Just try it out to see that your assumption is wrong. Perhaps you are referring to the current practice of all AI providers to censor what may be shown and how things should be framed. That can't be helped, I'm afraid, as long as there are only "western" AI providers who are trying hard to impose their world views (aka bias) on the rest of the world. Perplexity, I _guess_, does something similar with their search results by means of super/system/admin prompts to make the answers politically correct.
Tbh I prefer Krawl AI over Perplexity. It's more work-focused and I can get lot more things done using it because of the "tools" concept they introduced. Plus they are adding new tools to it almost every other week. It's not necessarily great for scientific research yet, but still more useful than Perplexity in other use cases I think
If you ask a question, but the answer only lies in a RU-vid video, can it find that? Or if you upload a csv file, can you ask it to find what the revenue for last month was?
Perplexity does read the transcript of RU-vid videos. Gor documents it can read PFFs but not sure about csvs, for that chatgpt is ptobably still the best option
Genini is like ChatGPT. Its a language model. Perplexity can use Gemini behind the scenes to summarize info across different links but it wont do the same thing
Never forget Tversky 💯, Tversky would be getting his nobel alongside Kahneman if he hadn't passed away so young :( This is a great, informative video. Thank you for sharing.
The way we think, act & react accordingly. So, let's know how to frame a right mental model. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-_LjjqYILwmI.html
