Passer au playerPasser au contenu principal
#ParlonsIA CHAT GPT - Perplexity -Claude
  • il y a 1 an
https://youtu.be/do8SlpMaL7c

1 like de la video = 1 merci ❤️ MES FORMATIONS → https://parlonsia.teachizy.fr/ On se demande si l'usage intensif de chat gpt et de l' artificial intelligence a un réel impact de l'ia sur nos capacités cognitives. Cette vidéo explore comment la technologie, y compris chatgpt 5, influence notre manière de penser et d'apprendre avec gpt5, la révolution got-5cognitive.


🌐 https://parlonsia.teachizy.fr
📺 https://www.youtube.com/@IAExpliquee.x
📺 https://dailymotion.com/formation.ai87
📘 Facebook : https://bit.ly/4kabhuA
🐦 Twitter / X : https://x.com/ParlonsIAx
📩 Contact : formation.ai87@gmail.com
🎙️Podcast – https://spoti.fi/4dqZ3uO
✍ BLog: https://medium.com/@flma1349/
💃https://www.tiktok.com/@parlonsia

---------------------------
IA a tester :

Coder agent IA: https://bit.ly/Coder_agentiA
Short AI: http://bit.ly/4lzE782
SEO agent IA : https://urlr.me/P8AS5N code rection 25% : PARLONSIA25

⌚ Timeline Vidéo
0:00 : ChatGPT vous rend IDIOT ?
1:20 : L'analogie du TSUNAMI
2:45 : J'ai reçu de NOUVELLES MENACES
4:40 : On va chasser les LICORNES
6:10 : utomatisation BTP
9:55 : 160 000 vues, 0 esprit critique le déclin cognitif.
11:50 : coder avec Perplexity Lab ?
15:10 : Le PROMPT PARFAIT ?
18:00 :GPT-5 Mes Astuces.



gpt5,chatgpt 5,ai,gpt 5 coding,gpt 5 review,chatgpt,openai,ai news,gpt-5,chatgpt users,l'ia rend bete,chatgpt nous rend il plus bete,ai trends,chatgpt feedback,prompt engineering,prompter chatgpt,prompt parfait,chatgpt coding,chatgpt tutorial,agi,gpt 5,ai updates,chatgpt prompt,GPT 5,GPT5

Catégorie

🤖
Technologie
Transcription
00:00Hello, Cortés, have we all become idiots? Have we all become stupid from using TchatGPT?
00:06In any case, that's what the MIT study would tend to say, TchatGPT takes it into account.
00:10So I asked myself the question, how did you learn to use artificial intelligence?
00:15Where did you see your first tutorial? And do you suffer from this condition, cognitive decline?
00:21So, like most people, in my opinion, it's with the big social media influencers.
00:26Genoa isn't so much about Ayo Miyazaki getting robbed.
00:29Ayo Miyazaki? Ayo! Ayo! Miyazaki! Fratello!
00:33Are you still scribbling on the breast?
00:34Whether you are afraid or not, reality will always remain the same.
00:38Okay, are you ready for the most fallacious comparison of the century?
00:40When a tsunami comes, far away, you cannot negotiate with the tsunami.
00:44We can't tell the tsunami, please come back next week.
00:47We can't call politicians and say, please ban the tsunami.
00:51In the face of a tsunami, the only thing we can do,
00:53it is to go to an area located 30 meters above sea level,
00:56or 3 kilometers inland.
00:57It's the surfer.
00:58Oh yeah! Oh yeah! Oh yeah! Oh yeah! The jerk!
01:01But how?
01:02It's about putting yourself in the corner and trying not to be crushed by the force.
01:05Oh yeah! Oh yeah!
01:06Oh yeah! Fratello!
01:08As long as you're in the corner, it's fine.
01:09I asked Lya for the image of a tsunami, so the angle is there.
01:12For in real life, he is there.
01:13Or there. Or there.
01:14So, we all learned from these experts,
01:17and today we will continue to study these experts,
01:20what they told us about automation, coding and the perfect prompt.
01:24And I have to tell you that I have received more threats,
01:26It's not the first time, but this time we're going to talk about it.
01:28I am told very clearly that if I continue to analyze their content,
01:32something will happen to me, and they will be less accommodating.
01:35They also boast of having a like rate of over 96% across the entire channel.
01:41Wow!
01:41If guys have so many likes, it must be quality.
01:45And that I absolutely should not attack channels like Defend Intelligence or OpenAI,
01:49and that, I assure you, I identified who the author of this email is,
01:53and that he tells me about a certain Ludovic who had already suggested that he attack me.
01:57So, I'm sorry to tell you, I will continue to analyze your account.
02:00Yeah!
02:00Because in France, we have one thing, it's called freedom of expression,
02:03and it is the intellectual property code, article 122-5, paragraph 3,
02:07CHR freedom of expression, article 10,
02:10and if not, you have the option, quite simply, of applying article 29-18-80.
02:16So, when I saw, indeed, the posts of this famous Ludovic
02:20which asks its community of more than 250,000 people,
02:24where nearly 26,000 people were reached by his message,
02:27where he indicates being attacked by a narcissistic pervert.
02:29On the other hand, using these methods where we manipulate people to make them believe that we are victims,
02:34This is rather large-scale manipulation.
02:37So, you understand, I'm not going to have fun studying this character here,
02:40who is very well known on social networks,
02:42and who, following the video Prompt GPT-5 Perfect, here is how to do it,
02:47does not wish to debate the content,
02:49but rather highlight that it is certified by Microsoft and OpenAI,
02:53the only French person to obtain certifications,
02:55invited by BFM TV, one of the best news channels, by the way,
02:59and he worked for GAFAM and unicorns.
03:01So, today, I suggest we go hunting for unicorns,
03:04in automation, in code and perfect ompt.
03:08Now we're going to see a video where there are more than 160,000 views on this content,
03:12on GPT chat automation.
03:14I'll give you the full prompt so you can do your tests too.
03:18I always told you, what to do,
03:20This is not taking literally what these great specialists tell us.
03:23invited to Vivatech and BFM TV,
03:25how these great specialists built an automaton in the construction sector.
03:30So, if you are thinking of scaling your business,
03:34you know, in all the training of these great YouTubers,
03:38They explain to us that we can automate GPT chat,
03:40we can automate GPT chat,
03:42and we will see how they proceed.
03:44So we have a system that will automate GPT chat
03:47to make quotes in the construction world,
03:49to encrypt your hardware,
03:51to tell you how much material to use for an instruction.
03:54In short, you have a complete system.
03:56So, I'll show you the full prompt.
03:57I am a B2B sales professional,
03:59I'll let you take a screenshot so you can use it,
04:02and that will give you an idea of how they code system instructions.
04:07So, I'll read you right away,
04:08I don't do it like that at all, not at all.
04:10So, to introduce myself,
04:11I am one of the first French speakers, certainly,
04:13which has entered a number of LLMs, if you like,
04:16and since the beginning of my training,
04:18I was inspired by one thing,
04:20prompt systems,
04:21how companies code their own instructions.
04:24So, I have never done in my classes, differently than I see here,
04:28instructions of this type.
04:29Never.
04:30But nevertheless, what I want to reveal to you is the result.
04:33Looking at a glance,
04:34and if you are among the 160,000 people who have seen it,
04:37what shocked me,
04:38is that out of 160,000 people,
04:4096% of people think it's good.
04:42In the idea in question,
04:43we want to build a pergola,
04:45and so we ask the AI to build a solid pergola,
04:4910 meters by 10 meters,
04:50and in the way of prompting,
04:51There are things that are, for me, extremely embarrassing.
04:54We want to beautify a garden with a beautiful, solid pergola,
04:56and not just any pergola,
04:58but one that will be the sensation of the neighborhood.
05:00I think Chad GPT doesn't care about the neighborhood feel.
05:02We are talking about a pergola made of this good cedar wood.
05:06I don't know if Chad GPT has any business with it,
05:08whether it is cedar wood.
05:09And she wants it to feel like a little getaway
05:11to the outside of his house.
05:13So, from there,
05:14when these great specialists,
05:15always invited to hunt unicorns,
05:18and on BFM TV,
05:20will justify that in their tutorial,
05:22viewed more than 160,000 times,
05:23that I would have preferred it to be removed,
05:25in view of the danger that this presents,
05:27he receives a response from the AI telling him
05:29“We are going to build a pergola measuring 10 meters by 10 meters,
05:32it's already a 3-car garage,
05:34with a height of 6 meters.
05:35So it's a 2 story.
05:37Have you seen many 2-story pergolas?
05:38With 20 centimeter poles,
05:40with 15 wood screws and metal spacers.
05:43We've all seen metal gaps
05:45and 15 wood screws.
05:46200.
05:46200 is enough.
05:47And without bolts, galvanized nuts.
05:49What is embarrassing,
05:50It's just that I don't see the dimension.
05:52Brief.
05:53Well, if you are in construction,
05:54I think you got it.
05:55Never mess around with AI.
05:58And what is terrible,
05:59it's that no one has tilted
06:00saying to themselves, "Hey guys,
06:0110 meters by 10 meters,
06:026 meters high,
06:03with advice like that."
06:05And AI, artificial intelligence,
06:07she is aligned.
06:08Aligned with "I'm going to have to answer
06:10to the user question.
06:12And so she's completely screwed up.
06:14She doesn't have the skills
06:15to respond to problems.
06:16And the very clear question that arises,
06:17It's when these professionals
06:18teach you that
06:19in mainstream videos?
06:21What do they teach you?
06:22in the classes?
06:23That's the question I ask myself.
06:24So tomorrow,
06:25They don't agree.
06:26Well,
06:26they will present to the judge
06:28that AI is capable
06:29to define from this prompt
06:31a pergola structure
06:33and that it does not present any danger
06:34with a height of 6 meters
06:36and that when you click
06:37on the product reference,
06:38building platform
06:39and King Merlin,
06:40you automatically have the prices.
06:42I tested it.
06:42You have the prompt,
06:43you copied and pasted it,
06:44you test it
06:44and you will see that
06:45nor the prices,
06:46nor the structure
06:47are accessible
06:48and should absolutely not be useful.
06:50So, it's a matter of common sense.
06:51and that's when I tell you,
06:52MIT neuroscientists
06:53are categorically used
06:54of GPT cat,
06:55It makes you stupid.
06:56And they noticed anyway,
06:57but still,
06:58than 93% of 18-25 year olds
07:00use GPT chat,
07:01that's everyone.
07:02And 50% for everyday use.
07:04The problem,
07:05This is when we use GPT chat,
07:06we deactivate almost 65%
07:09of our cognitive capacity.
07:11We don't think anymore.
07:12So when you have
07:12160,000 people
07:14who see content
07:15and who don't tilt on it,
07:17That raises questions.
07:18Warning, alert.
07:19What to expect
07:20like skidding
07:21so that we become aware
07:22errors and dangers
07:24tutorials
07:24what are we giving to AI today?
07:26If we put as only word
07:27this type of content
07:28and that we do not oppose
07:30to this type of content,
07:31I assure you,
07:32you will not benefit from anything
07:34of this type of interaction.
07:35On the contrary,
07:35you put yourself in danger
07:36especially if we apply it
07:37in the same way.
07:38Now that your intellectual chakra
07:41began to open,
07:43It's time to move on.
07:44But that's crazy,
07:45I did a random moment
07:46of the video, guys.
07:47I have testimonies
07:49of people who worked
07:50to Facebook
07:50in the first 20 employees.
07:52Zuckerberg,
07:52he's a guy
07:53who enters a meeting
07:54and he says the illegal,
07:55button there,
07:56you have to put it on the left.
07:56No, but that's crazy,
07:59I did a random moment
08:00of the video, guys.
08:01But how will you be
08:03credible, what?
08:04That's kind of what I tell myself,
08:06how do they arrive
08:07to be credible, anyway?
08:08So, just for the record,
08:10for these clever little ones,
08:11these AI specialists,
08:13know that I have already filed
08:14files,
08:15so all the contents
08:16which were blown up,
08:17instructions are already underway.
08:19So, at one point,
08:20there is someone who will pay.
08:21Because if article 122-3,
08:24INEA 5 is well respected,
08:26at one point,
08:27there is someone
08:27who will go to the checkout.
08:29So, keep having fun.
08:30But you see,
08:31complaints,
08:31they are already filed
08:32and I will have the information.
08:33So, we will continue to analyze
08:35and we're not going to stop there.
08:36This time,
08:37we will analyze a part
08:38where we are told
08:39that we are going to code an application
08:41with the brand new tool
08:43Perplexity Lab.
08:45I made a video of it,
08:45a video,
08:46I studied
08:479 AI expert training courses.
08:49Not one,
08:509.
08:50You have them in description,
08:51you can watch the video
08:52and I made a training course out of it
08:53hyper accessible
08:54to answer
08:55to all these problems
08:56that I noted.
08:57Act like an expert
08:57AI industry analysis.
08:59He must do,
09:00I'll let you copy the prompt,
09:01you will test this prompt.
09:03If you have a Perplexity account,
09:05you can do it
09:05and if you don't have one,
09:07I gave you a tip
09:08to have Perplexity
09:09at 3 euros per month.
09:11It's in the description in the video,
09:123 euros per month.
09:13I pay annually
09:14less than 27 euros,
09:15I believe,
09:16the 1-year pro subscription.
09:17All this to tell you
09:18that in this test here,
09:19we say,
09:20always the specialists,
09:21the great unicorn hunters,
09:23guests still on BFM
09:24and VivaTech,
09:25AI must aggregate
09:26AI trends,
09:27compile trending topics
09:29on AI,
09:30on Twitter,
09:31Reddit and LinkedIn
09:32over a 30-day period.
09:34There too,
09:34I say to myself,
09:35wait,
09:35there are 30,000 people
09:36who see a video
09:37and nobody tilts.
09:39Do you see
09:40where is the problem?
09:41SO,
09:41it's really in the sentence
09:43in red which is here.
09:43I will reveal to you
09:44my point of view
09:45on the question.
09:46Guys,
09:46you all saw
09:47than on the chains
09:48vision style thing
09:50and the others,
09:51they are already on their way
09:52to talk about GPT-6
09:53while they are never
09:54arrived to show us
09:55consistent use
09:56of artificial intelligence
09:58in my opinion.
09:58When you are offered
09:59this type of content
10:00and that no one tilts
10:02saying to himself,
10:03can we really,
10:04with a prompt like this,
10:06aggregate 30 days
10:08of data,
10:09from Twitter,
10:09from Reddit
10:10and LinkedIn?
10:11Is anyone
10:12asked this question?
10:13I did it.
10:14So, we show you
10:14on screen
10:15an amazing app
10:16where you click
10:17and surprisingly,
10:18you have LinkedIn,
10:19Twitter, Reddit
10:20and the whole analysis,
10:22the whole analysis,
10:23I tell you.
10:23You can turn
10:24all your teams
10:25data strategist,
10:26analysts,
10:27marketing,
10:28Short,
10:29you have everything,
10:30All,
10:30everything that is done.
10:31At home,
10:31it works in one click,
10:33It's incredible.
10:33And that's why,
10:34Besides,
10:34maybe we invite them
10:35on TV sets.
10:36There was a time,
10:36it was Patrick Sébastien
10:38for the cabaret
10:39or, I remember,
10:40Sunday,
10:41the circus,
10:41for those who remember
10:42still from that time.
10:43And yes,
10:44Today,
10:45I will show you
10:45that I had appraised
10:47the number of pages
10:48by Perplexity
10:49that corresponds
10:50one month of data.
10:51SO,
10:52one month of data,
10:53it corresponds approximately
10:55to 90 million posts,
10:57approximately,
10:57for Reddit only.
10:59SO,
10:59according to the information
11:00that I was able to harvest,
11:01the interpretations
11:02the most plausible
11:03are between 22 million
11:05to 26 million per month
11:07only for Reddit.
11:08SO,
11:09you imagine
11:09than artificial intelligence
11:11will process in 15 minutes,
11:13I'm talking about this interface,
11:14the famous interface
11:15of Perplexity
11:16that I tested,
11:16that I put in the course
11:17and I'll explain to you
11:18at the limit
11:18how we treat
11:19this type of problem.
11:20It is that beyond
11:21pretty graphics
11:22that we take out for you,
11:23can you imagine
11:24that the AI has just processed
11:2626 million pages
11:28just for Reddit?
11:29Just for Reddit.
11:30Because you have
11:31Twitter, Reddit and LinkedIn.
11:33I'll call you back anyway.
11:33the specialist's pedigree
11:35who worked
11:36within GAFA
11:37and tech unicorns.
11:38So, still,
11:38he has a notion
11:39of this type of information.
11:41But in the mail
11:42that I received,
11:43I am told elsewhere
11:44that I am nobody.
11:45I came out of nowhere
11:46and that I have no right to criticize.
11:48Otherwise, it will happen to me
11:49misfortune.
11:49But I continued
11:50to investigate
11:51because I said to myself
11:52but in my opinion,
11:53Twitter and LinkedIn
11:54are not accessible.
11:56And that's what I have
11:56actually controlled
11:57with AI.
11:58There is no possibility
11:59to make an access
12:01to the proprietary platform.
12:03So all the data
12:04which are shown
12:05as extracted data
12:07social networks
12:08and categorization
12:09platforms
12:10are most likely false.
12:12LinkedIn is a platform
12:13owner
12:13with strict rules
12:14API and scraping.
12:16So, that surprised me.
12:17that we could have shot
12:1822 million in 15 minutes
12:19and Twitter is the same.
12:21So you realize,
12:22there are still 30,000 people
12:23who have seen this content,
12:24who have seen this type of prompt
12:26and who may have
12:27learned this type of method
12:28and who said to themselves
12:29it's great
12:30from now on.
12:31I just have to ask the AI
12:32to be an expert analyst.
12:33I ask him
12:34and boom,
12:35It just falls on the table.
12:36That is until the moment
12:37where you believe in unicorns.
12:38But when you no longer believe in unicorns,
12:40you realize
12:41that these methods,
12:42these methods which are shown,
12:43they are viral.
12:44Everyone's getting involved
12:45on the networks saying
12:46everyone can do it.
12:47I mean,
12:48everyone is capable
12:48to write this type of sentence.
12:50But the problem,
12:50It's only when you realize
12:52that it is absolutely unusable
12:54in the professional world
12:55and that people like me
12:56tell you,
12:57we find it right away
12:58less sexy.
12:59And we realize one thing,
13:00It takes skills.
13:01Perhaps it is necessary
13:01stop the bullshit.
13:02And that's when you realize
13:03that AI,
13:04it may be
13:04the biggest speculative bubble
13:06of the last ten years
13:08in which 95%
13:10projects are abandoned.
13:12But today,
13:12why we can't get there anymore
13:13to backtrack
13:15in marketing game?
13:16That's what it represents
13:1772% today
13:18of American investment.
13:20I tell you very clearly,
13:21if on social networks,
13:22we were beginning to become aware
13:24that we cannot do
13:2580% of what is claimed
13:27in the tutorials,
13:28there would be a collapse
13:29markets.
13:30And in my opinion,
13:31we can ask ourselves the question
13:32if one day,
13:32it couldn't happen
13:33if he doesn't arrive
13:35to hide this bitter failure.
13:37SO,
13:37is it necessary
13:38drop
13:38artificial intelligence?
13:40No.
13:40There really is
13:41big advantages
13:42to use AI,
13:43but to use it
13:43with the right methods,
13:44you will actually
13:45save time.
13:46And that's what I show
13:47in training.
13:48We won't have viral things
13:49as we have just seen here,
13:50but we will have reliable things.
13:52SO,
13:52what I suggest you do,
13:53it's now
13:54to study the part
13:55of the perfect prompt
13:56and see the use
13:58of GPT-5
13:58which is very misunderstood
14:00for a very simple reason.
14:01They spent 3 years
14:02to show you
14:03tutorials
14:03that don't work
14:04and interactions
14:05which are not good.
14:06Don't be surprised
14:07Today,
14:07Unfortunately,
14:08to have grown up,
14:10fed
14:10with this type of content.
14:11Once you have
14:12became aware of this
14:13and specialists
14:14who showed you
14:14this type of content,
14:15Today,
14:16Me,
14:16I want to
14:16they say
14:17I attack you,
14:18do it.
14:18You go before the judge
14:19and you show him
14:20the results of your codes,
14:22the results
14:22of data extraction,
14:23the result
14:24of automation
14:25and you show
14:25to the judge
14:26whether yes or no
14:27what you have shown,
14:28that works.
14:29And if it doesn't work,
14:30you risk passing
14:31a little for...
14:31Finally,
14:32you can always say
14:33that it is the fault
14:33to GPT chat.
14:34In any case,
14:34This is what the study says.
14:35Come on,
14:35we will look now
14:36the perfect prompt.
14:37I made a video
14:37completely dedicated
14:38at the perfect prompt,
14:39you have it in description
14:40if you want to watch it,
14:42but we will touch
14:43new topics
14:43to make you understand
14:44how to use it.
14:45First element,
14:46there is the arrival
14:47in the tutorial I saw,
14:49I had the link for you
14:49from the original tutorial,
14:51the arrival of variables.
14:52Earness,
14:53Level,
14:53Clarification,
14:54Ask and Autonomy.
14:55So all that,
14:56In my opinion,
14:56it's pure invention
14:58which does not exist.
14:59In addition,
14:59we have one thing,
15:00is that you actually have
15:01a code here
15:02which is well in place
15:02which changes over time.
15:04We have a form of XML
15:05with mixed Markdown,
15:07that's how it is
15:07that's what it's called,
15:08and then we have a shape
15:09from Markdown
15:09with typos.
15:10So the person
15:11who shows you this tutorial
15:12and who is a great specialist
15:13from the field of AI,
15:14eminently known
15:16and recognized
15:16to hunt the unicorn,
15:18it's not for him anyway
15:19appeared in the tutorial.
15:21And me,
15:21it still raises questions
15:22how we arrive
15:24to let pass
15:24similar breads
15:26and how we get there
15:27to get out of sentences
15:27which do not come out
15:28of any tutorial.
15:29It is enough,
15:30I showed it to you
15:30in the other tutorial,
15:31to control
15:32HLGPT,
15:32and have the reflex
15:33to ask him
15:34is there any
15:35official documentation
15:36on this content,
15:38no results,
15:39the documentation
15:39does not evoke any element,
15:41the only thing
15:42that he shows,
15:43This is the concept
15:43of persistence.
15:44And that,
15:44it's interesting
15:45and it's there
15:46where the solution lies.
15:47What he calls
15:47AI autonomy
15:49is defined
15:51in the book
15:51Open AI Book.
15:53These are things
15:53that you learn
15:54in training.
15:55As I explained to you,
15:56I'm leaving
15:57prompt systems,
15:58I'm talking about methods
15:59official work
16:01which are used,
16:02I study them
16:03and I show you
16:04what works
16:05and what doesn't work.
16:05In these methods,
16:06there is a section
16:07who is called
16:08Persistence.
16:08I translated it for you.
16:09in French
16:10to make you understand
16:10how do we create
16:11of autonomy
16:12in a system
16:13said agentic.
16:14Here is the prompt,
16:15you can capture it
16:15and use it
16:16in your interface.
16:17As an agent,
16:18you must not go
16:19until resolution
16:20complete request,
16:21only ends once
16:22the problem solved
16:23with certainty,
16:24in case of uncertainty,
16:25adopts the approach
16:25the most logical
16:26and continues
16:26without interruption.
16:28Never wait
16:28of human validation,
16:30make assumptions
16:30the most reasonable,
16:31act then write it down
16:32for the user
16:33after execution.
16:33These are instructions
16:34which allow
16:35to develop
16:36indeed autonomy
16:37of the system
16:37by giving him
16:38very clear instructions.
16:39And these are the instructions
16:40of open AI management
16:41which I have nevertheless optimized.
16:43There they would be raw.
16:44We also have
16:44agentic elements,
16:46that's to say,
16:47the agent,
16:47it's that the model
16:48is able to follow
16:49instructions.
16:50Its neural system
16:51go further
16:52than simple prediction.
16:54These are systems
16:54word prediction,
16:55the LLMs.
16:56But today,
16:56they are capable
16:57to keep lines
16:58driving.
16:59So if you want,
17:00we will see examples
17:01to make you understand better,
17:02give to AI
17:03the possibility
17:04to search
17:05on the Internet,
17:06you can give it to him
17:07For example
17:07the number of times
17:08that he can connect,
17:09the amount of data
17:10that he can integrate
17:11and sorting
17:12that he must do
17:13of this information.
17:14And that,
17:14it's called
17:15a gathering context.
17:16So how do I know all this?
17:17I told you,
17:18I made tutorials
17:19and in training,
17:20there it is
17:20Prompt Engineering Advanced
17:21where I enter the systems.
17:23So, when I got back
17:23in systems,
17:24you learn something,
17:25how are they written
17:26the instructions
17:27real LLMs,
17:28not stuff
17:29that you find there,
17:30on the tutorials
17:31specialists
17:32unicorn hunters
17:33on the Internet.
17:34From there,
17:35In fact,
17:35in training,
17:36we study in detail
17:37something
17:38which can be
17:39more difficult
17:40for the majority of people.
17:41And that,
17:41it's the training
17:42Prompt Engineering Excellence,
17:44the structure
17:45for systems
17:46Prompt Systems,
17:48AI Agents,
17:48and Chatbot.
17:49Today,
17:49what will interest me,
17:50it is to give you
17:50solutions
17:51and these are solutions
17:52that I share
17:53in training
17:54The Best of AI
17:55which is a training,
17:56in my opinion,
17:56that I recommend to you.
17:57Already because
17:57it is the most complete,
17:58certainly extremely honest
18:00on the capacities
18:00models.
18:01We don't study just one model,
18:02but we study almost
18:03all models on the market.
18:04In any case,
18:04you have the methods
18:06that you can use
18:06both for the office,
18:08for the image
18:08and for the video.
18:09NOW,
18:10I will show you
18:10how we will use
18:13ChatGPT
18:13by making interactions
18:15that work.
18:16You don't have it at all
18:17need to structure
18:18the information
18:19of this style
18:20with the code
18:21in pseudo-XML.
18:23I will show you
18:24methods
18:24that work
18:25with an example.
18:26When you ask
18:26information
18:27which the model has,
18:28the date of the First World War,
18:30if you add
18:31the text variable,
18:32you add
18:33the verbosity variable.
18:35Verbosity,
18:36it will define
18:37the length
18:38of the model's response.
18:39There are three types
18:39of lengths,
18:40low, medium and high.
18:42High,
18:42it's a system
18:43who will give
18:44more depth
18:45and examples.
18:46If you need
18:47to increase what is called
18:48reasoning,
18:49you will see one thing.
18:50The reasoning,
18:51reasoning effort,
18:53to trigger it properly,
18:54the advice I give you,
18:56you saw something,
18:56than in the previous case,
18:58we have no reasoning
18:59which was triggered,
19:00otherwise we would have
19:00a little cone of reasoning.
19:01That I put
19:02reasoning effort
19:02and it didn't go off.
19:04The way,
19:05and I give you this tip,
19:06which allows you to trigger well
19:07the reasoning system,
19:08It is to use the word use.
19:10Use reasoning
19:11high effort
19:12and in this case,
19:13you have the reasoning
19:17automatically
19:17in model thinking
19:19from an automatic model.
19:20And there,
19:20you have decomposition
19:22of the operation of the model.
19:23SO,
19:24this little word,
19:24worn,
19:25reasoning effort,
19:26tends to be better understood
19:28by the model
19:28that the sentence
19:29reasoning effort.
19:30I will show you
19:30an example now
19:32where we will mix
19:34functions
19:35of webtools,
19:36so these are tools
19:37that we will trigger
19:37of the system,
19:38only with words.
19:40You saw,
19:40we do not activate any element,
19:42but we will use
19:42systems
19:43who are called
19:44the tools,
19:45the tools,
19:45and you will tell me
19:46"But where do you come from?
19:47these elements?
19:48Once again,
19:49I take them out of the system prompt.
19:50When you integrate
19:51the information
19:52from the system prompt,
19:53you access
19:54to the structure
19:55of these tools
19:56and you understand better
19:57how you can
19:58call them.
19:58SO,
19:58the idea there,
19:59that's what he's going to have
20:00a medium reasoning
20:02in research
20:03of the answer,
20:04a relatively extensive text
20:06using
20:07just before
20:08to answer
20:09an internet connection
20:10and he will introduce me
20:11his answer
20:12directly in Canva.
20:13SO,
20:14if you want to use
20:14this same system,
20:15you are able
20:16Today
20:17either to click
20:18on the buttons
20:18to be able to do
20:19this action.
20:19But what becomes
20:20interesting,
20:21this is when you have to
20:22develop what we call
20:23agentic systems.
20:24When you have
20:25GPTs,
20:26you can now
20:27integrate into the instructions
20:29behaviors
20:30of the model
20:30depending on the steps.
20:32And that makes the model
20:32much more efficient
20:33and efficient.
20:34Shall we take this test together?
20:35Come on, let's launch.
20:36The model should therefore
20:37start thinking
20:38and it is.
20:39We'll see what he's going to do
20:40search for information
20:41in its database.
20:41Afterwards,
20:42it should connect
20:43also to the Internet.
20:44We will access
20:45to his system of reasoning
20:47and he should finish
20:48by the fact
20:49that he will look for
20:50sources on the Internet
20:51and present
20:52the entirety
20:53of its content
20:54on the web.
20:54If you see
20:55than a tool
20:56does not trigger,
20:57don't hesitate
20:58to put
20:58the little word
20:59worn
20:59just before
21:00and in general,
21:01it improves
21:02the model's response.
21:03SO,
21:03you see that now,
21:04we didn't click,
21:05we didn't check
21:06of elements,
21:06but we have the possibility
21:07to modify directly
21:09in the Canva interface
21:10and it allows you
21:11to work
21:11in localized
21:12on the account.
21:13SO,
21:13this system,
21:14you can completely
21:14import it
21:15in GPTs
21:17and therefore allow
21:18to your machines
21:19to integrate
21:20this type of behavior
21:21in function
21:22of the stage
21:22that you are in the process
21:23to develop.
21:23SO,
21:24be very clear,
21:24prompt engineering,
21:26it's not an option.
21:27It certainly is
21:27Today
21:28that architecture
21:29which is based
21:30on what we call
21:31transformers
21:32is the tool
21:33number one
21:34to be able to work
21:35and understanding
21:36really thorough
21:37of these tools
21:38is essential.
21:39If we continue
21:40In fact
21:40to believe
21:41that it is enough
21:41to say
21:42you have an assistant
21:42specialized
21:43in strategic analysis
21:44to think
21:45that we will obtain
21:46a content
21:46level
21:47strategic analysis,
21:48we are wrong.
21:49Working methods,
21:50you understood,
21:51In fact,
21:51are very far away
21:52from what we are shown
21:53on social networks.
21:54It is certainly
21:54much less,
21:55I would say,
21:56bankable
21:56because it does
21:57less flashy
21:58but there is
21:59real strategies
21:59so that people
22:00don't be disappointed,
22:01don't waste time
22:02and don't lose either
22:03money
22:04in content,
22:04training courses.
22:05Those who really have
22:06need today
22:06to work
22:07and save time
22:07have a great tool,
22:09you have to learn
22:10to use it
22:10and that's what I propose
22:11to do.
22:12So I won't stop
22:13to analyze the contents,
22:14to test what they say
22:15these great specialists
22:17of the AI field
22:17and if they have to answer,
22:18they just have to answer
22:19by showing that what they do
22:20in the tutorials
22:21works.
22:22And for the moment,
22:22three tutorials,
22:23three things that don't work
22:24and which are not valid.
22:25I find that
22:26to be referenced
22:27on GAFA
22:29and the unicorn,
22:30fortunately that
22:31from time to time,
22:31we test to see
22:33how reliable it is
22:33what we are told
22:34on the network.
22:35Please feel free to share
22:35en masse on Twitter,
22:37on Facebook,
22:38on LinkedIn.
22:39This video,
22:40this is the best way
22:41to create a community
22:42who is fed up with marketing
22:44and who wants it today
22:45to learn to use
22:45correctly the AI.
22:46Come on, I tell you
22:47see you soon,
22:47see you next time.
22:48French subtitles

Recommandations