Thank you for using Julius and spreading the word, Andy! I'm the founder of Julius and I've been a huge fan of your channel since before Julius was even built so seeing this video means a lot ❤ Julius AI recently crossed 600k users in a short amount of time. Hoping to reach millions of academics and PhD students soon 🙏
Hey! I'm just wondering if Julius is using ADA (Advanced Data Analysis) as its engine, or if it's something different. Excellent job, by the way! Thanks! Also, what about privacy data? Can we send the CSV file of the bank's most wanted billionaires? :)
@@joseluisgonzalezgeraldo1577 Thank you 🙏 We don't use ADA as our engine, we have built up our own infrastructure which gives us far more flexibility. For example, you can now do R based analysis in Julius instead of needing to learn R Studio. It also gives us flexibility to work with best models out there like GPT-4, Claude-3, Llama-3 70B, Command-R, Gemini 1.5, Mistral etc. Soon we're adding GPUs in a partnership with Nvidia to provide accelerated computations for some researchers too.
Something really worth addressing is the ethics of using these apps. Universities and similar institutions will probably strongly advise against uploading your data to a third party application. Who owns the data after you've uploaded it? Will your (still unpublished) data be used to train new models? And a whole host of other problems that prevents us from safely using these tools beyond the 'toy' phase.
Hi, none of the data uploaded to Julius or your prompts are used towards future model training. We utilize strict access controls and have strict data retention policy. Every dataset uploaded to Julius gets deleted after 1 hour of inactivity, no datasets are retained. Let me know if you have more questions.
@@rahulson23 good to know. However, my remark was much more general, as Andy discusses all sorts of AI tools and they all have their own terms of service and these terms can change at any time and...I mean, it's just the same as with everybody who created content on the internet and now sees their content used by OpenAI.
@@rahulson23 but you do know that data does not fully disappear. Its always recoverable and reusable by the model itself in future searchs. You just have to ask the correct questions and youll get that data back. All of us know this so...
I have tried the free version. To extract data.. however, it hasn’t work a single time. I can buy it. But someone normally wants to be sure that the car works, so you don’t need 4 persons to push it. I am probably stupid but it’s the first app I found protesting.. any thoughts?
I was prepared to get my Jupyter notebook going before voyaging into Julius but since it already uses my favorite libraries. Maybe I will skip that part now. This is great. I do love my jupyter stuff and using cool new libraries for insights but this is game changer.
as a researcher you must approach research tools with critics if they are good mention it, if not same and please compare same type of tools with each other
Very useful as usual. Can I please ask how do you decide which AI model to use? I have had perplexity for a year and it has been reasonable for a variety of things. How do you decide on the best AI model for phD research? There are so many. I find myself trying all of them and ending up wondering what I'm doing 3 hours later, without making any progress.
Hey just found your video, dont suppose you know if it can integrate with Google Sheets and/or Looker studio? Im a marketer and thinking how to create a dashboard that can show CEOs key metrics but also have Julius show insights and recommendations on how to improve campaigns. 🤔
Yes, Julius can create R code. I never used the R code. I have used it for connecting to another SaaS via API and Julius wrote the Python code for analysis, executing some functions on our Saas platform account. It would taken us probably a week or more but I could finish the work in one night using Julius.
Sure thing, just export it to .csv. But be aware of data policy! maybe you are sending hot stuff you should not (what about data privacy?). In fact I have done something similar with ADA (former Code Interpreter) in ChatGPT IV and works just superb. Waiting to be published, by the way.
PLEASE DO A VIDEO ON FULL PROFESSORS WHO TAKE FIRST-AUTHORSHIPS IN PHD AND POSTDOC STUDENTS' WORK. THIS IS A HUGE SCAM IN ACADEMIA AND RESEARCH WHICH IS TOTALLY HUSHED UP AND UNADDRESSED.
@@boredscientist5756 I'm in such in a situation as an international student that if I do that, he'll sue my PhD and my 4 years would be wasted. Great if someone can set an example by doing that !!! At the very least, it would uncover this totally hushed up academic and research scam.
That makes no sense. You mean understanding text? AI can do that better than nvivo. But mostly “qualitative” data is already quantitative because survey responses have been codified for example.
Produce novel research? Aka the actual objective of the doctoral student? Frankly, it's not like many PI's teach the PhD student how to do the stats, yet it's an unavoidable expectation nowadays. For example, my PI during my PhD had no idea how to do stats but expected results using R. Ironically, during their PhD, they hired a statistician to do their basic [descriptive] stats, which was the norm then. It is strikingly obvious that expectations have shot up for doctoral students in the last few decades, both in the level of stats required and in producing it independently regardless of the PI's level of expertise with statistics. Services like these help make specialised expectations accessible and actually give a doctoral student, for example, time to focus on producing quality research. Rather than spend months just learning how to open a working directory and import their data. With that, just because the program can spit out code, does not mean it's accurate and frankly, it still warrants a level of understanding of A. your data and B. the test and its output. The assumption that these new programs nullify that entire branch of the research process during a doctorate appears to be based on misunderstanding and stubbornness to recognise the potential benefits, only because it's a new and massive technological development. Not assuming this is your point of view, but this stubbornness that is crocking up between us faculty also tends to contribute to the mentality "I did my PhD this way and suffered, so everyone else must too or else their doctorate does not hold the same merit". We don't have to automatically embrace every AI program, but assuming something like this discredits what a PhD student still has to do along the way is an incredibly silly comment to make.
This looks interesting. I have what I think should be an easy task for AI, but nothing is yet capable. I need AI to look at two uploaded PDFs: one with over 100 questions and the other containing information needed to answer those questions. I also need AI to identify questions in a PDF and convert the questions into a txt file that can be uploaded to the Canvas LMS. So far, nothing.
I think no commercial AI can do what you are asking for. AI (Large Language Models) are mere superpowerful statistical predictive machines. They work really well when you have structured and clean data like in this video. In you case, the task is more complex. Although there are some AI out there which can read pdfs and answer questions based on them, it is not that easy for them because of the structure of the pdf where you have a mixture of headers, titles, text, images, tables, footnotes, etc. And even more complex when you have two pdfs. Moreover, AI apps are not the same, some of them are trained more towards instructions (you give specific instructions to perform something), others are trained in conversations or chats, others are trained to produce code, others are trained to extract text, etc. You would need something that can extract and identify structures of pdfs and also have the functionaity of transforming file types. I can recommend you to explore LMstudio and AnythingLLM ua-cam.com/video/-Rs8-M-xBFI/v-deo.html
@@datagus Great response. Thanks. Yes, even ChatGPT4 can't currently identify the internal structures of a PDF well enough to do what I want it to do. ChatPDF, even the free version, comes close, but I have to convert the Questions PDF to txt and feed the AI questions manually with a certain character limit.
@@pequodexpress There are some algorithms and perhaps some apps outthere that can transform the pdf into text. In your case, what you are asking that one algorithm breakdown the structure of your pdf, identifies the structure of format of those questions, and transforms them into anyother file type. That can be done without AI with some algorithms using python. You can include AI in those algorithms and the response will be better. For your case, you would need a customized AI-based algorithm. You need to ask a Data Scientist or an AI/ML engineer for this.
Hi Andy. Your content is always useful but you have become something of a parody of yourself, of your communicative intent and idiomatic style. The game of UA-cam is not the game of PhD...
Thank you for using Julius and spreading the word, Andy! I'm the founder of Julius and I've been a huge fan of your channel since before Julius was even built so seeing this video means a lot ❤
Julius AI recently crossed 600k users in a short amount of time. Hoping to reach millions of academics and PhD students soon 🙏
Can PhD students get a discount code from you
Hey! I'm just wondering if Julius is using ADA (Advanced Data Analysis) as its engine, or if it's something different. Excellent job, by the way! Thanks! Also, what about privacy data? Can we send the CSV file of the bank's most wanted billionaires? :)
@@nicholasasante2295 they can, i just got one
@@joseluisgonzalezgeraldo1577 Thank you 🙏 We don't use ADA as our engine, we have built up our own infrastructure which gives us far more flexibility. For example, you can now do R based analysis in Julius instead of needing to learn R Studio. It also gives us flexibility to work with best models out there like GPT-4, Claude-3, Llama-3 70B, Command-R, Gemini 1.5, Mistral etc. Soon we're adding GPUs in a partnership with Nvidia to provide accelerated computations for some researchers too.
I am definitely going to use it, for sure thank you very much.
Something really worth addressing is the ethics of using these apps. Universities and similar institutions will probably strongly advise against uploading your data to a third party application. Who owns the data after you've uploaded it? Will your (still unpublished) data be used to train new models? And a whole host of other problems that prevents us from safely using these tools beyond the 'toy' phase.
Hi, none of the data uploaded to Julius or your prompts are used towards future model training. We utilize strict access controls and have strict data retention policy. Every dataset uploaded to Julius gets deleted after 1 hour of inactivity, no datasets are retained. Let me know if you have more questions.
@@rahulson23 good to know. However, my remark was much more general, as Andy discusses all sorts of AI tools and they all have their own terms of service and these terms can change at any time and...I mean, it's just the same as with everybody who created content on the internet and now sees their content used by OpenAI.
@@rahulson23 yes, we'll take your word for it 😅
@@rahulson23
@@rahulson23 but you do know that data does not fully disappear. Its always recoverable and reusable by the model itself in future searchs. You just have to ask the correct questions and youll get that data back. All of us know this so...
I have been using it for a week now and am obsessed!
I have tried the free version. To extract data.. however, it hasn’t work a single time. I can buy it. But someone normally wants to be sure that the car works, so you don’t need 4 persons to push it. I am probably stupid but it’s the first app I found protesting.. any thoughts?
This was so helpful thank you!! Saved me hours of work. Am a new Julius subscriber now 💪
I was prepared to get my Jupyter notebook going before voyaging into Julius but since it already uses my favorite libraries. Maybe I will skip that part now. This is great. I do love my jupyter stuff and using cool new libraries for insights but this is game changer.
Thanks for sharing this. I hope it helps with molecular dynamics analysis
as a researcher you must approach research tools with critics if they are good mention it, if not same and please compare same type of tools with each other
That means less videos...
@@leondbleondb fewer
@@tuckerbugeater 👌
the new interface of julius is much more intimidating :)
Really great. Thanks for sharing this wonderful information
Got the julius annual subscription. Costly but useful 👍
Data scientist in the pocket should be the thumbnail! :D
Very useful as usual. Can I please ask how do you decide which AI model to use? I have had perplexity for a year and it has been reasonable for a variety of things. How do you decide on the best AI model for phD research? There are so many. I find myself trying all of them and ending up wondering what I'm doing 3 hours later, without making any progress.
Hey just found your video, dont suppose you know if it can integrate with Google Sheets and/or Looker studio? Im a marketer and thinking how to create a dashboard that can show CEOs key metrics but also have Julius show insights and recommendations on how to improve campaigns. 🤔
I love your hoodie! Where did you get it from? :)
Does it have the capability of data cleaning and processing before getting into the visual?
Its amazing must say
Can you link the RICE thing you were talking about?
Can you do a walk through of analyzing something then writing the results section as per a journal regulations?
Please
Instead of Python code, can it produce R code for code books and markdown documents?
Wondering the same thing
Yes, Julius can create R code. I never used the R code. I have used it for connecting to another SaaS via API and Julius wrote the Python code for analysis, executing some functions on our Saas platform account. It would taken us probably a week or more but I could finish the work in one night using Julius.
Can I input my SSPS data?
Sure thing, just export it to .csv. But be aware of data policy! maybe you are sending hot stuff you should not (what about data privacy?). In fact I have done something similar with ADA (former Code Interpreter) in ChatGPT IV and works just superb. Waiting to be published, by the way.
Is there any ai tool for analysing molecular data, something like structure, darwin but in AI form
Just so incredible 👍🏾👍🏾
Can this tool be used for qualitative data analysis to generate codes and themes?
That's what I am using it for. Currently in the midst of a thematic analysis of social media posts.
what video recording tool are you using with the colorful background that moves.. ? loom? Thx
That is added in during editing.
Hi andy would you suggest any course for understanding research writing
Another ❤
Can it analyse spss data?
Just export the data to another format, FFS
Legend!
Is it free or requires a license? How is the data protection? I am wondering because the data we are uploading is original/unpublished
It has a free version but he is using the paid one
Do you think the free version is good enough?
Its limited
"delve" 9:36
?
So much GPT outputs in his verb diet XD
PLEASE DO A VIDEO ON FULL PROFESSORS WHO TAKE FIRST-AUTHORSHIPS IN PHD AND POSTDOC STUDENTS' WORK. THIS IS A HUGE SCAM IN ACADEMIA AND RESEARCH WHICH IS TOTALLY HUSHED UP AND UNADDRESSED.
You can literally sue the full Pr. for that
@@boredscientist5756 I'm in such in a situation as an international student that if I do that, he'll sue my PhD and my 4 years would be wasted. Great if someone can set an example by doing that !!! At the very least, it would uncover this totally hushed up academic and research scam.
@@boredscientist5756no you cant
@@boredscientist5756how could you justify an amount to sue for?
Stop crying.
Just remember to be transparent in your research by stating that you generated the graphs with this software.
Bad idea
I’m wondering how Julius AI would do with qualitative data… Nvivo would still be our main bet.
That makes no sense. You mean understanding text? AI can do that better than nvivo. But mostly “qualitative” data is already quantitative because survey responses have been codified for example.
Great. So what does the doctoral student do?
Produce novel research? Aka the actual objective of the doctoral student? Frankly, it's not like many PI's teach the PhD student how to do the stats, yet it's an unavoidable expectation nowadays. For example, my PI during my PhD had no idea how to do stats but expected results using R. Ironically, during their PhD, they hired a statistician to do their basic [descriptive] stats, which was the norm then. It is strikingly obvious that expectations have shot up for doctoral students in the last few decades, both in the level of stats required and in producing it independently regardless of the PI's level of expertise with statistics.
Services like these help make specialised expectations accessible and actually give a doctoral student, for example, time to focus on producing quality research. Rather than spend months just learning how to open a working directory and import their data. With that, just because the program can spit out code, does not mean it's accurate and frankly, it still warrants a level of understanding of A. your data and B. the test and its output. The assumption that these new programs nullify that entire branch of the research process during a doctorate appears to be based on misunderstanding and stubbornness to recognise the potential benefits, only because it's a new and massive technological development. Not assuming this is your point of view, but this stubbornness that is crocking up between us faculty also tends to contribute to the mentality "I did my PhD this way and suffered, so everyone else must too or else their doctorate does not hold the same merit". We don't have to automatically embrace every AI program, but assuming something like this discredits what a PhD student still has to do along the way is an incredibly silly comment to make.
This looks interesting. I have what I think should be an easy task for AI, but nothing is yet capable. I need AI to look at two uploaded PDFs: one with over 100 questions and the other containing information needed to answer those questions.
I also need AI to identify questions in a PDF and convert the questions into a txt file that can be uploaded to the Canvas LMS.
So far, nothing.
I think no commercial AI can do what you are asking for. AI (Large Language Models) are mere superpowerful statistical predictive machines. They work really well when you have structured and clean data like in this video. In you case, the task is more complex. Although there are some AI out there which can read pdfs and answer questions based on them, it is not that easy for them because of the structure of the pdf where you have a mixture of headers, titles, text, images, tables, footnotes, etc. And even more complex when you have two pdfs.
Moreover, AI apps are not the same, some of them are trained more towards instructions (you give specific instructions to perform something), others are trained in conversations or chats, others are trained to produce code, others are trained to extract text, etc. You would need something that can extract and identify structures of pdfs and also have the functionaity of transforming file types.
I can recommend you to explore LMstudio and AnythingLLM
ua-cam.com/video/-Rs8-M-xBFI/v-deo.html
@@datagus Great response. Thanks. Yes, even ChatGPT4 can't currently identify the internal structures of a PDF well enough to do what I want it to do. ChatPDF, even the free version, comes close, but I have to convert the Questions PDF to txt and feed the AI questions manually with a certain character limit.
@@pequodexpress There are some algorithms and perhaps some apps outthere that can transform the pdf into text. In your case, what you are asking that one algorithm breakdown the structure of your pdf, identifies the structure of format of those questions, and transforms them into anyother file type. That can be done without AI with some algorithms using python. You can include AI in those algorithms and the response will be better. For your case, you would need a customized AI-based algorithm. You need to ask a Data Scientist or an AI/ML engineer for this.
Try Claude 3 (Opus is best recommended), Gemini 1.5 or Perplexity (pro model best recommended)
Am just guessing though
@@timooothy1234 Thanks for your reply. Your suggestion sounds interesting.
Is there any Beaty and nice man discount?😂??
You sound like totally manic .... but that is because I am playing this 2x speed :) anyhow very clear introduction thanks
Hi Andy. Your content is always useful but you have become something of a parody of yourself, of your communicative intent and idiomatic style. The game of UA-cam is not the game of PhD...
Huh?
What Do you mean?
I think this would make sense if it made sense