Perplexity is so awesome. Im using it to explore hypothesis and dig into different research topics, it's so useful and just great to use :) Thanks again for another awesome video!
But I just clicked on your TH-cam account and it is obvious that you are a bot. Aside for this comment you have zero activity what-so-ever. Thank you, mr. Bot.
Perplexity was amazing a year ago, it's been seriously under performing for a while though, many times not following even the clearest of instructions. I wish they'd fix it already.
It also seems to lose track into a long thread, even repeating itself. I find that sometimes I have to take the sub-sub-sub topic I discovered and start a new thread with it.
I am happy to see that other people are reporting the same thing. Last time I mentioned this in the comment section I was accused of secretly working for OpenAI.
Perplexity is a nice tool. It finds real documents, and generally limits its answers to those sources. This means there is much less room for hallucination. Its built on other LLMs, so it can go off on a tangent with its own knowledge. when it does do, its less reliable. Be aware. Look to see if it goes beyond its listed sources. It will fill in some details - from the listed sources - with its own knowledge. this is ok if they are just minor details. For example, it uses a standard easy to read style, regardless of the style of its sources. I find that it relieves me of tedious research, and cuts thru links that go on forever. You might say that Perplexity is the first true "knowledge engine," rather than a "weight / parameter based engine." For most LLMs, we dont have a way to verify that wts are correct, other than crude back prop and gradient descent - which is very crude but all we have - short of Perplexity. One day all parameters in LLms will shift to real knowledge bits - Perhaps rules or semantic nodes by the millions or billions. They can be checked for accuracy, one at a time. So Perplexity is a precursor to the day when LLMs will contain real knowledge. Its main functions: o it finds documents o It puts them in a standard format summary - that you can check against the found docs. Easily. o It does not put the main points in subordinate clauses - or imitate the crappy writing habits of some authors. o its an intelligence multiplier for humans. o We can focus on what we are good at. Einstein: "imagination is greater than knowledge." Use our vision, dont plod thru shit work. However Dont think that AI will get smarter and smarter. There is speed limit to AI, its like c, the speed of light. Call it a. Its caused by a math wall and other factors. When you accelerate close to c, energy requirements go to infinity. Same with a. Energy needs become monstrous and impossible. Would it take all the energy in the solar system to achieve and IQ of 1 million? lacking the imagination of Einstein, would it still be below humans?
I have been using claude sonnet to do the same. And it is freakingly good. I want to compare it to perplexity. Hopefully, I will get the courage to let go another 20 dollars.
Awesome tool! Very nice, even though its prime for ocums razor scientific standard reduction of big bang computed/space between . Any day etymology will simulate convergences its just a Matter of compute lol
I find perplexity is far.more useful than chatgpt but i actually find the academic mode unusable, when i run it on literature that i know in depth it usually fails to find top references and pver half are completely different fields!
I feel like Perplexity, just like the models it uses, is pretty good at giving general information, but isn't great with details (I know models can make things up, but still…). Undergrads and grad students might find it handy for a quick overview of a subject, but it's not very helpful for brainstorming advanced (biomedical) science, yet. Or it’s just my ignorance…
Just avoid telling your old school supervisor, otherwise he will ask you to go to the library and search by printed journal issues 😂. Doesn't matter his field even if he works in latest technology he will resist and we will have Martin Luther situation or even John Wycliffe 🔥 and you will get excommunicated 😉 😂
You’ve got to stop doing this, man. I really appreciate the educational content in your videos, but the comedy bits are getting to me! I’ll be paying close attention, fully focused, and then you throw in that cow bit, and I end up laughing so hard I’m on the floor. I keep rewatching it, and it’s tough to get back into the educational part. So go easy on that mate! 🤣🤣
Can you explain how AI can help to "OUTDATE" publications? We suffer in many fields from thousands of papers which are not valid any more because science has proven them to be wrong. How do we ever get rid of such papers. (Example: corrupt HIV and Climate Hoax claims)
Research papers are horrible to read? Sir. Someone making a well-reasoned argument with data is quite enjoyable to read. I like how your videos are almost all facts and very little opinion. You just show processes, what things can do. Also, research papers can sometimes contain the best inside jokes almost no one will be able to appreciate. Chef's kiss.
Once again, YOU are advocating for Big Tech... WE are supposed to trust servers run by other companies just to get information that makes our lives easier...
I think it's wonderful that when you ask Perplexities about the problem it has with its unethical use of copyright material - it tells you all about how bad they are with that and if you ask if it might threaten the future of the company, they'll tell you that too. It was fun to play around with questions about their intellectual theft and unethical business practices. Perhaps they're the next AltaVista
perplexity ai is a weak platform in the face of other ai's that are frankly useless. I don't understand why people advertise it, don't waste your money on it.
this also misses important details in paper . Sorry I had to unsubscribe from you Andy because you are promoting irresponsible AI. please just stick to basic phd advice and not LLMs.
![Things you don't say out loud in academia [9 open secrets]](http://i.ytimg.com/vi/sgAlFPMugWw/mqdefault.jpg)
![Things you don't say out loud in academia [9 open secrets]](/img/tr.png)
![[#2024MAMA] G-DRAGON - 무제(Untitled, 2014)+POWER+HOME SWEET HOME+뱅뱅뱅+FANTASTIC BABY | Mnet 241123 방송](http://i.ytimg.com/vi/Ox29z5Nf1Uk/mqdefault.jpg)
Perplexity is so awesome. Im using it to explore hypothesis and dig into different research topics, it's so useful and just great to use :)
Thanks again for another awesome video!
But I just clicked on your TH-cam account and it is obvious that you are a bot. Aside for this comment you have zero activity what-so-ever. Thank you, mr. Bot.
@@ValSchnitzel what?
Definitely. I been using perplexity for a while now. What model do you prefer when using Perplexity AI and why?
Pro settings and citations are LIT 🔥
Good timing! I was looking for a paper to prove a point and scite and google scholar came up with zero.
Perplexity was amazing a year ago, it's been seriously under performing for a while though, many times not following even the clearest of instructions. I wish they'd fix it already.
It also seems to lose track into a long thread, even repeating itself. I find that sometimes I have to take the sub-sub-sub topic I discovered and start a new thread with it.
Last time I tried Perplexity the performance was pretty poor. I wouldn't pay for it.
I agree, I've been using it for more than an year, it's performance has definitely gone down
I am happy to see that other people are reporting the same thing. Last time I mentioned this in the comment section I was accused of secretly working for OpenAI.
That's weird, it's just gotten better for me and my team over time
Perplexity is a nice tool. It finds real documents, and generally limits its answers to those sources. This means there is much less room for hallucination.
Its built on other LLMs, so it can go off on a tangent with its own knowledge. when it does do, its less reliable. Be aware. Look to see if it goes beyond its listed sources.
It will fill in some details - from the listed sources - with its own knowledge. this is ok if they are just minor details. For example, it uses a standard easy to read style, regardless of the style of its sources.
I find that it relieves me of tedious research, and cuts thru links that go on forever.
You might say that Perplexity is the first true "knowledge engine," rather than a "weight / parameter based engine." For most LLMs, we dont have a way to verify that wts are correct, other than crude back prop and gradient descent - which is very crude but all we have - short of Perplexity.
One day all parameters in LLms will shift to real knowledge bits - Perhaps rules or semantic nodes by the millions or billions. They can be checked for accuracy, one at a time.
So Perplexity is a precursor to the day when LLMs will contain real knowledge. Its main functions:
o it finds documents
o It puts them in a standard format summary - that you can check against the found docs. Easily.
o It does not put the main points in subordinate clauses - or imitate the crappy writing habits of some authors.
o its an intelligence multiplier for humans.
o We can focus on what we are good at. Einstein: "imagination is greater than knowledge." Use our vision, dont plod thru shit work.
However
Dont think that AI will get smarter and smarter. There is speed limit to AI, its like c, the speed of light. Call it a. Its caused by a math wall and other factors. When you accelerate close to c, energy requirements go to infinity. Same with a. Energy needs become monstrous and impossible.
Would it take all the energy in the solar system to achieve and IQ of 1 million? lacking the imagination of Einstein, would it still be below humans?
Perplexity is simply the best AI tool out there.Absolutely amazing what the team has done
They should reach out to you... Because of you I just got the pro.
If you pay for perplexity you get more pro searches and can change the AI Model.
I wish I had discovered this before I started writing my thesis, but at least its still early.
when you say keep saying delve into , I got more feeling that your video content is also made by one of the ai tools you used often :D
@@username42 All my own content. Just like saying that word for some reason.
@@DrAndyStapleton So does ChatGPT
Getting data only from sources like semantic scholar is a big downside, as it can't access scopus or sciencedirect.
I'm a Perplexity Pro user for medical research, and it has rendered sources from ScienceDirect many times in my searches.
its very flawed.
Andy, a nice sense of humor. Is perplexity paying you to make these videos?
I wish!
@@DrAndyStapleton They should. You are doing a great job explaining Perplexity. Good luck!
I have been using claude sonnet to do the same. And it is freakingly good. I want to compare it to perplexity. Hopefully, I will get the courage to let go another 20 dollars.
I think you can ask it to provide the output without bullet points 🙂
Awesome tool! Very nice, even though its prime for ocums razor scientific standard reduction of big bang computed/space between .
Any day etymology will simulate convergences its just a Matter of compute lol
Loveeee perplexity!!!
Hi Andy! Is there a trustworthy website or platform for attending upcoming workshops and conferences?
2:56 Three, sire.
I find perplexity is far.more useful than chatgpt but i actually find the academic mode unusable, when i run it on literature that i know in depth it usually fails to find top references and pver half are completely different fields!
I had the same experience. Perplexity is giving me incorrect references.
Just ask it to rewrite your answer without bullet points. Easy.
Connected Papers are brilliant. But only two graph per month. Is there any alternatives?
what about research rabbit or litmaps?
@@DrAndyStapleton Research Rabbit is equally good.
I feel like Perplexity, just like the models it uses, is pretty good at giving general information, but isn't great with details (I know models can make things up, but still…). Undergrads and grad students might find it handy for a quick overview of a subject, but it's not very helpful for brainstorming advanced (biomedical) science, yet. Or it’s just my ignorance…
Just avoid telling your old school supervisor, otherwise he will ask you to go to the library and search by printed journal issues 😂.
Doesn't matter his field even if he works in latest technology he will resist and we will have Martin Luther situation or even John Wycliffe 🔥 and you will get excommunicated 😉 😂
You’ve got to stop doing this, man. I really appreciate the educational content in your videos, but the comedy bits are getting to me! I’ll be paying close attention, fully focused, and then you throw in that cow bit, and I end up laughing so hard I’m on the floor. I keep rewatching it, and it’s tough to get back into the educational part. So go easy on that mate! 🤣🤣
Hi Andy . It work in spanish?
I’m not Andy but yes all the AI programs work in Spanish
Can you explain how AI can help to "OUTDATE" publications?
We suffer in many fields from thousands of papers which are not valid any more because science has proven them to be wrong. How do we ever get rid of such papers. (Example: corrupt HIV and Climate Hoax claims)
is it more powerful than chat GPT 4 ?
I'm a Pro user. You can choose Chat GPT4 as your LLM and Perplexity searches the web, which ChatGPT4 can't do.
Research papers are horrible to read? Sir. Someone making a well-reasoned argument with data is quite enjoyable to read. I like how your videos are almost all facts and very little opinion. You just show processes, what things can do. Also, research papers can sometimes contain the best inside jokes almost no one will be able to appreciate. Chef's kiss.
Chatgpt sub is more than enough
still has issues.
Once again, YOU are advocating for Big Tech...
WE are supposed to trust servers run by other companies just to get information that makes our lives easier...
drag and drop, haha
Delve again! Noooooooo. GePeTo, get out of that body!
Your mistake „thas“ instead of „this“ is actually correct in Bavarian German 😂 Des (pronounced like thas) means this or also the
I think it's wonderful that when you ask Perplexities about the problem it has with its unethical use of copyright material - it tells you all about how bad they are with that and if you ask if it might threaten the future of the company, they'll tell you that too. It was fun to play around with questions about their intellectual theft and unethical business practices. Perhaps they're the next AltaVista
perplexity ai is a weak platform in the face of other ai's that are frankly useless. I don't understand why people advertise it, don't waste your money on it.
Dont, I spent and regretted in 10 mins in
Perplexity is terrible. Most of the time the answer you get is wrong. I stopped using it. This seems like a paid ad video.
YOU should stop starting your videos with the Tripe 6 hand sign - it's just too OBVIOUS!
Ad
Troll.
its an ad
this also misses important details in paper . Sorry I had to unsubscribe from you Andy because you are promoting irresponsible AI. please just stick to basic phd advice and not LLMs.