Prompt engineering is dead but there is a better life after death.
Summary
TLDRIn this insightful presentation from AI 5050, Bob Alink challenges traditional prompt engineering methods, advocating for a deeper understanding of AI's capabilities. He introduces 'prology,' a blend of technical and psychological knowledge, to master AI prompts effectively. Alink emphasizes the importance of asking AI questions to avoid assumptions and achieve non-average results. He showcases creative prompting techniques, such as using regional nuances and psychological frameworks, to harness generative AI's full potential, ultimately urging professionals to delve deeper into AI models for exceptional outcomes.
Takeaways
- 🔄 The presenter, Bob Alink, emphasizes the need to 'turn the world upside down' and look at AI from a different perspective to avoid average results.
- 🤖 He criticizes the current approach to 'prompt engineering' in 2024, likening it to the outdated practices of 1924, where people tried to extract information and enforce their will on others.
- 📚 Bob dismisses cheat sheets and shortcuts as ineffective, suggesting that they are based on assumptions rather than a true understanding of AI's capabilities.
- 🧑💼 He points out the misuse of roles and frameworks in AI, where people assign roles to AI without fully understanding what those roles entail or how to effectively use them.
- 🔍 Bob introduces the concept of 'prology', a blend of prompt engineering and psychology, as a way to deeply understand and interact with AI models.
- 📝 He advocates for asking AI models questions about their processes, rather than just giving commands, to gain insights into how they interpret and act on prompts.
- 🌐 The script discusses the importance of specifying regional contexts when prompting AI to avoid stereotypes and to tailor responses appropriately.
- 🌈 Bob highlights the potential of generative AI to create unique combinations of ideas and techniques that have not been previously combined, emphasizing the role of human creativity in this process.
- 📈 He provides examples of how to use AI for specific tasks, such as creating brand voice profiles or adapting text for low-literacy audiences, by asking the model to explain its actions.
- 🚀 The presenter encourages the audience to think creatively and to combine different psychological models and techniques in AI to achieve better results.
- 🛠️ Bob acknowledges the challenges of evolving AI models, which require ongoing learning and adaptation of prompting techniques to maintain effectiveness.
Q & A
What is the main theme of Bob Alink's presentation at AI 5050?
-The main theme of Bob Alink's presentation is exploring new perspectives on how AI is influencing the world and the concept of 'prompt engineering', which he believes is flawed and needs to be rethought.
Why did Bob Alink start his presentation standing upside down?
-Bob Alink started his presentation standing upside down to symbolize the idea that if AI is turning the world upside down, he needs to look at it from a different, less conventional perspective.
What does Bob Alink think about the use of cheat sheets and shortcuts in prompt engineering?
-Bob Alink considers cheat sheets and shortcuts to be worthless, as they are based on assumptions and do not truly reflect the capabilities or the workings of AI.
What is Bob Alink's opinion on the use of roles in AI prompting?
-Bob Alink believes that using roles in AI prompting without understanding what the template or role actually does can lead to assumptions and ineffective results.
What does Bob Alink propose as an alternative to traditional prompt engineering?
-Bob Alink proposes 'prology', a combination of prompt engineering and prompt psychology, as an alternative approach to better understand and utilize AI.
What does Bob Alink mean by 'average is the new AI'?
-By 'average is the new AI', Bob Alink is referring to the phenomenon where AI, when used by everyone in the same way, tends to produce average results, rather than exceptional or unique outcomes.
What technique does Bob Alink suggest to gain deeper insights into AI's understanding and use of prompts?
-Bob Alink suggests using the 'what will you do' technique, where instead of executing a prompt, one asks the AI to explain what it would do with the prompt, to gain deeper insights.
How does Bob Alink approach the issue of AI's stereotypical responses?
-Bob Alink addresses AI's stereotypical responses by explicitly instructing the AI to avoid stereotypes and to consider regional and cultural nuances in its responses.
Outlines
🔥 Challenging AI Prompt Engineering Norms
Bob Alink discusses the limitations of traditional AI prompt engineering, comparing the current state to the 1920s where people sought to extract and enforce information. He criticizes the use of cheat sheets and assumptions in prompting, advocating for a deeper understanding of AI's capabilities. Bob introduces the concept of 'prology', a blend of technical knowledge and communication psychology, to master AI prompts effectively. He emphasizes the need to ask questions and understand AI's true actions to escape average results and achieve exceptional outcomes.
🗣️ The Art of Talking to AI: Prology
This paragraph delves into the concept of prology, where Bob explains the importance of communicating with AI effectively. He suggests that the way we prompt AI now is flawed, as we often assume the AI understands our intentions without verifying. Bob introduces the idea of asking AI questions to understand its processes better, which he believes is crucial for learning how AI interprets and acts on prompts. He shares his personal journey of discovery with AI, emphasizing the need for creativity and logical sense in understanding AI's behavior.
🌐 Cultural Nuances in AI Prompting
Bob highlights the importance of considering cultural and regional nuances when prompting AI, using the example of how AI defaults to an American accent despite being prompted in different languages. He suggests that specifying the region can lead to more accurate and culturally sensitive responses. Bob also touches on the use of roles in prompts, recommending that including the country with the role can significantly improve the AI's output by avoiding stereotypes and tailoring the response to the specific cultural context.
📝 Exploring the Depth of AI Prompt Interpretation
In this paragraph, Bob explores the intricacies of how AI interprets different words in prompts, such as 'generate', 'make', 'create', and 'craft'. He demonstrates that these words can elicit different responses from the AI, depending on the context. Bob encourages users to experiment with various words and ask the AI to explain the differences in its actions. This approach can lead to a deeper understanding of AI behavior and, ultimately, more effective prompting.
🎨 Breaking Away from Average AI Results
Bob discusses the concept of 'average' AI results and how to transcend them through deeper prompting and creativity. He emphasizes that as AI evolves, it learns from the average user, which can limit its potential for creative and unique outputs. To counter this, Bob suggests using more specific and creative prompts, combining elements that have never been combined before, which he refers to as generative AI. This approach requires human creativity and can lead to more colorful and diverse AI outputs.
🤝 Combining Psychological Models for Enhanced AI Outputs
This paragraph focuses on the use of psychological marketing frameworks to enhance AI outputs. Bob mentions several psychological models and principles, such as Cini's principles and the FAB model, and suggests that combining these can lead to more effective AI responses. He also discusses the importance of asking the AI about the feasibility and effects of combining different models, which can result in unique and powerful AI applications.
🛠️ Mastering AI with Creative and Analytical Prompting
Bob shares his experience in creating detailed voice and style profiles for AI using up to 80 elements, which allows him to control the output of any text. He discusses the challenges and time investment required to master AI prompting, emphasizing the need for creativity, logical and analytical thinking. Bob also touches on the evolution of AI models and the necessity to continually adapt and deepen one's understanding of AI to maintain control over its outputs.
🚀 The Evolution of AI and Continuous Learning
In the final paragraph, Bob addresses the ongoing evolution of AI and the need for continuous learning and adaptation. He mentions the emergence of new AI models and the potential for previous prompts to become less effective over time. Bob encourages the use of creative thinking and the development of new prompts that combine novel elements. He also highlights the potential of AI agents and the importance of personal involvement in creating and optimizing prompts to achieve exceptional results.
Mindmap
Keywords
💡Prompt Engineering
💡Average Results
💡Roles
💡Frameworks
💡Prolology
💡Generative AI
💡False Friends
💡Non-Stereotypical
💡Curiosity Gap
💡Marketing Psychology
💡Evolution
Highlights
The presenter, Bob Alink, emphasizes the need to look at AI from a different perspective, suggesting that traditional prompt engineering is insufficient.
Alink criticizes the common practice of prompt engineering, comparing cheat sheets and assumptions to worthless shortcuts.
He introduces the concept of 'prology', a blend of prompt engineering and psychology, to better understand AI's response to prompts.
Alink advocates for more in-depth questioning of AI models to grasp how they interpret and execute prompts.
The idea that AI models can be influenced by specifying regions or countries to avoid stereotypical responses is discussed.
Alink demonstrates how different wording in prompts, such as 'generate' vs. 'craft', can lead to distinct AI behaviors.
The presenter suggests that generative AI can produce more creative and less average results by combining elements in new ways.
Alink explains how to use AI for tasks like creating low-literacy friendly texts and avoiding 'false friends' in different languages.
The importance of combining psychological marketing frameworks with AI to achieve targeted outcomes is highlighted.
Alink showcases the creation of detailed voice and style profiles using up to 80 elements for precise AI text control.
The evolution of AI models is addressed, noting that as they develop, professionals must adapt their prompting strategies.
The presenter discusses the use of 'what will you do' technique to gain insights into AI's interpretation of prompts.
Alink illustrates the power of three-word prompts in marketing, leveraging psychological principles for impact.
He emphasizes the need for creativity and deep understanding to harness generative AI's full potential.
Alink provides examples of how to prompt AI for non-stereotypical outputs by specifying regional nuances.
The concept of using AI as a tool for optimization, rather than relying solely on its generated prompts, is introduced.
The presenter concludes by encouraging continuous learning and adaptation in the ever-evolving field of AI prompting.
Transcripts
hi my name is Bob alink and this is a
recording from the presentation I gave
at AI
5050 few weeks ago in
Brussels view differently New
Perspectives does AI turn the world
upside down I have already started now
you there is also a reason why I'm
standing upside down because if I AI
turns the world upside down I have to
look at it upside down which very little
people are doing and that's what this
presentation is
about PT engineering to me in
2024 it looks like
1924 where we wanted to extract
information out of people we want to
enforce them our will and that's how
most people
prompt we determine or we say we want
this and we just want the
AI to listen to us we want it to obey
that's how we prompt engineer in
2024 and all of you no cheat sheets no
yapping offer money my life depends on
this I use it as toilet paper to me it's
worthless
crap
it's shortcuts which don't work
assumptions
and the best toilet paper to me is the
ones which claim 10 times better
performance to me this is like a 10
layer toilet people it's very soft on my
well you know
what another thing most people are using
you are now act as you are now a
copywriter with 10 years experience and
blah blah blah blah blah you're a
marketeer and again putting the model in
roles is putting the model in a template
but you have no clue what the template
is about again you assume
this and what I like most is what I call
impressive prompts when you read the
prompts and the models the roles they
are so impressive that as a human you
might think well this should give an
incredible result copyr with 10 years
experience and also marketeer your SEO
expert and 10 rolls in one uh all better
than the other ones uh it's you
don't know what a role does no one knows
but you can ask that's what I will tell
you later in this
presentation the same applies to
Frameworks we use
Frameworks and I'm not against
Frameworks they're very nice but before
you start building a
framework you need to know what you're
building and what materials you're
building it with again now you're using
roles you use words which you assume do
something you put them in a
framework and then you hope it'll
work framework should be something solid
like in this image of Steel you should
know what you're building with you
should know what the prompts really do
not
assume only when you master what prompts
really do you can start building
Frameworks and my opinion 99% of all the
people doing prompting have no idea what
their prompts are really
doing and what does this give average
results now average isn't bad but
average is average and what we see is
still more average average results and I
call it average
it's the new AI
average and the more you use Ai and the
more people use it in the way everyone
is using it now we will get average
results the trick is how to get out of
average and hopefully I'll can tell you
a little bit more about
that and because of this I think pumpt
engineering is dead the way we engineer
now
it's not the way forward so if you have
a phone now please look at it and turn
it into silence mode this is important
for the next
slide because
now we're going to remember pumpt
engineering so please show one
millisecond of respect to memorize pumpt
engineering but pum engineering is that
there is a better life after
death but what is this better
life pum engineering is that but there
is a better life after death and what is
this better
life let me tell you about
it the way we formed engineer now is
that so what we
start what we have to do is talk to the
model because
we use words now and we assume those
words do something it's like when you
talk to your colleague a friend
sometimes you only need to Blink a blink
of an eye or you need half a word and
that person knows what you
mean but when prompting you write down
words and you assume the AI does
something with those words you assume
what you think it does but it doesn't
always
so you have to start asking
it and how you can do that we'll come to
that later but we have to talk much more
to the AI what is doing the why how ask
questions questions questions questions
that's the only way you learn how the AI
really works with your
prompt and I call it it's I didn't make
up this word it's called prology
and if you do promy you become a
promist and of course I asked chat GPT
what is a
prologist
and the term suggests a blend of
technical knowledge like engineering and
an understanding of communication and
behavior AK psychology so
prology is a combination of prompt
engineer and a prompt psychologist
which is the best of both
worlds where does this lead to to boldly
go where no one has gone before because
that's where we are at the AI age we
assume but we really don't know what's
going on inside the
AI I hear experts making the AI models
and sometimes they don't even understand
why the model is doing what it
does so we have to start from scratch
and this reminds me when I was a little
kid my commodor
64 cassette tape and basically you had a
computer and there was no manuals you
copied software which was normal in
those days no manual you would just
start contrl a contrl b contrl Ctrl D
what does it do you have to
use logical sense
creativity to find out what does it do
no manual find it out
yourself and that's what I've been doing
the last 1 and a half years just walking
to the
model discovering things and it cost me
2,000 hours uh but it's very
interesting and what I've come up with
is what will you do
technique and what will you do will
actually ask the model what it does with
a prompt so what will you do when I ask
you to give the prompt and what is the
difference if I ask you to another
prompt do not execute the pump but
explain me what you will
do this has given me lots of insight in
what the model really does with a prompt
one of these insights
was we pumpt in English or German or I'm
Dutch uh we can pumpt in Portuguese
which I talked to someone recently
and most of you have seen that the model
sounds typical American so although we
pum in our own language the model still
sounds
American now the very simple solution
for that is tell the model which country
or region you're in and again that's
something you can ask the model what
will you do when they ask
you uh this is the region and without
region and then it will tell you you it
will go in a different mode and consider
the regional
aspects and again you can also ask it in
the roles what will you do when I ask
you to act as a copywriter to rewi the
text and what's the difference if I ask
you to act as a French Belgium
copywriter again this is an example
everyone uses roles now but before the
role you should really put the country
in front of it because putting country
in front of a role will put that role in
a more targeted uh targeted role for
your country you will notice it will get
better I will I'm again I'm not a fan of
copywriter or
marketeer uh there are different
techniques where I can control it much
deeper but if you use roles at least put
country in front of the RO it will make
the output uh a lot better and again you
can ask it here uh FR Belgium
copywriter uh what's the
difference ask it just use these
techniques it will give you more insight
in what it does now what I've also
found is it tends to get
stereotype like it gets stereotype
English it could get stereotype
Belgium again act as a non- stereotype
French Belgium
copywriter you will notice this and it's
like when you read the text you analyze
it you see well this is like stereotype
what you think put it in a prompt it's
as almost as simple as that and again
you can ask it ask it ask what you have
in your head ask it and compare it to
the prompt you were using
again this will give you Insight in what
the model does at least that's what I've
been doing the last nine months and it's
gotten me deeper and deeper in the
model understanding what it does is it
always right
no but n out of 10 it
is so copywriter General Improvement of
quality now if I took the French Belgium
copywriter it says faing the text to
French Belgium
speaking cultural and linguistic Norms
the non-stereotypical French Belgium
copywriter ensuring the Texas culturally
sensitive avoiding cliches and providing
a modern respectful portrayal of the
audience this is the summary of the
previous prompt I asked the model to
compare the three different
prompts it's as simple as that so if you
use non-stereo typical French Belgium
copywriter or
put Portuguese uh I talked with recently
talked to someone from Portugal and he
said well when I use Portuguese it
sounds uh almost Brazilian which in
Brazil they're using also Portuguese
turns out if you ask the model about
Portuguese it uses International
Portuguese so you really have to tell it
use Portuguese Portuguese targeted to
Portugal it sounds almost
ridiculous but this is how the model
works
what will you do
technique
uh I ask it
also uh generate make create or craft
curiosity Gap title now most of you like
I did in the past make this make that do
this do that generate make create or
craft all sound familiar
same but when you ask a model the
difference is it gets very
interesting when I use generate focuses
on a systematic perhaps more formulaic
approach make implies a bit more
customization based on provided details
create suggest a more original and
inventive process and craft emphasizes
precision and careful construction for
maximum effect so the next time you
are going to make a title or curiosity
Gap try craft or create you will find
that when you use craft a title the
model will use different parameters and
it will get a better you will get a
better result same for create now when
you make a Excel table for example you
could say generate make create or craft
it will won't make a difference again
you you can ask
it ask it when I make a table when I use
generate make Creator craft is there any
difference no because a table an Excel
table is an Excel table and it doesn't
make a difference but creating a text
crafting a text a title that makes a
difference again these are just four
words there are many more words which
you could use and ask about it so this
is what I mean with
you think you prompt something but you
have no idea what the model really does
and only by asking you can find out what
the model does with the words you're
using how does the model really use
those words in what mode will it use it
and again this will bring you
further and it will get you away from
average because aage that's what I see
in the new
models uh I have to
use deeper and deeper prompting to get
out of
aage people say the model gets better
yes it does for the average user using
less pumps it knows what you
mean but it will give you average result
based on what it learned from all the
people and it's taken the average
sponsors from those so yes for most
people it will get better but it will
get average and I call it
average what you really want is more
colorful you want to get away from the
average result as an expert in marketing
or copywriting or anything else in your
field you want to get away from the
average result you want more you want
better you want more colorful now here
we have some colors of green so if you
go and promp deeper you you will get
some nice shades of green there's a
little bit of blue but what you really
want colors you want all the colors you
want generative
Ai and generative AI is really
using the
model at its best it's combining things
which have never been combined
before uh an
example yenga block game yenga is is a
game now I tried a couple of months ago
or half a year ago use yanga as a
summary yenga is not a summary
technique and I asked the model well
what will you do when I use yenga as a
summary I don't know yenga summary
technique and I had to told well use
yenga like the block game as a summary
technique and then it starts a
generative approach
yenga is taking out
blocks and it says well I have to take
out pieces of text similar to
yenga until I have all the important
pieces of text it's a completely
different summary technique and that's a
good example of generative
Ai and when you use generative
AI your human creativity you can do the
most incredible creative things but in
the end it's all about your
creativity your creativity makes
generative AI possible the AI model
itself knows a lot but in the end it
doesn't know what it knows it only knows
what's combined in this data set so if I
combine like yenga and
sumary it's never been combined
before and that's generative AI so your
creativity this makes generative a i
possible and it will make it possible
that you outgrow the average the average
results a very nice example of that is
low literacy FSE friends
Nigeria I started with this on a project
a couple of months ago write a text for
low literacy which is a common problem
in the Netherlands and in many many
countries false friends I'd never heard
about it
but it's interesting because I ask you
well what will you do when I ask you to
write a text in low literacy and the
person I'm writing it for comes from
Nigeria then it came up with false
friends and false friends are basically
words in English which have a completely
different meaning in Nigerian or take
any other country so what you could do
is rewrite this
text for someone who has low literacy
is originally from
Nigeria and prevent false friends so
what it will do then is it will rewrite
this text in English or any other
language and it will substitute the
words false friend words for other words
so that people in this case from Nigeria
will not get confused now this is a very
nice example of generative
AI you need to be english- speaking and
from Nigeria to know these
words otherwise you would have no clue
about this so for me this is an eye
opener and again you can ask the model
about it what will it do uh is just
combining things which it already knows
it knows low literacy it knows false
friends and now you're going to combine
it it will not do it by itself unless
you tell it to and that's your strengths
as a
prologist your creativity you can do do
amazing things like this the courses for
this which you can follow TH th000 or
more and at the Press of a button using
a prompt you can use
this so again you can ask it what will
you do if when I ask you to rewi a text
for someone with low literacy in Belgium
was originally from Nigeria and prevent
the use of false friends explain what
you will do and now the model will
explain you what it will do
asking it's the only thing I do ask ask
compare only by asking and comparing you
will see the differences between small
words one word difference can make a
huge
difference compare ask and this is how I
learn and this is I think how you can
learn mastering the AI and get the best
results out of it yes it takes time it
takes a lot of time shitload of
time so there's no shortcut there's no
cheat sheets you have to
learn I can tell you now so you you can
learn from this and this will save you a
lot of time uh but in the end uh you
will have to master it and that takes
time another example marketing
psychology if you're into marketing F
Alini Pock caraman burgler David Rock
scar prevent reactants this is what I've
been using in my profiles that I built
and these are
all psychological marketing Frameworks
cini has seven principles he's the
world's best known influence specialist
BJ ful has a different system canaman
system one and two which says as a human
you look at things emotional which is
system one and rational you look at
technical specifications system two
berglar the dialogue method David Ro
scarf something completely different
again you can ask the model what is
David walk scarf model explain it to
me
these psychologists have never been
combined and the AI model knows them and
I can combine them that's again
generative AI you can combine all these
different psychological models but be
careful not all of these models can be
combined so you can again you can ask it
this is my
purpose this is my target audience I
have these and these psychological
principles uh can I combine
them and the model will tell you yes or
no I have used some psychological
systems and it would just say no it's
it's the opposite effect or it just
doesn't work you only know when you ask
the mother
and again it will give a pretty good
result uh when you ask
it nice example I heard about story
brand framework never heard of it before
it's like I I hear about things I never
heard before so the only thing I do is I
start asking the model what a story
brand framework and can I combine this
with this case the David Rock scarf
model which works very nice for human
resource for certain uh certain jobs
combination David Rock and story brand
is a killer not fall but again you can
ask
it again you hear you see different
things can I combine them now these are
Frameworks
from
marketing uh out of the box thinking is
really use the yenga technique or
summary which yenga and summary is
completely different so if you really
want to go out of the box uh use a
technique
from completely different market and use
it in your market
and combine it but ask it what it will
do only when asking it what it will do
you will know whether it works or not
give your purpose give your target
audience then you give the different
models Frameworks you want to use and
ask it can I combine this and what will
it
do you will get amazing results this way
but only when you ask it you will it
will tell you whether it works or
not it it might work because in the end
the AI just makes assumptions based on
what it
knows and sometimes it's completely
wrong they can be combined or the other
way around but nine out of
10 it's right so again it will make you
more creative if you're creative it will
make you more creative and you can do
things a lot
faster and one of the things I've used
in the past is the power of three words
in a row which many of you see in
marketing if you're marketing you will
know a fun restor effect which is
basically when I say I see a table a
chair a
lizard table now lizard doesn't fit into
this which is a from westof effect which
you will see on many web pages where you
see most salt it's it's highlighted now
you can do it subtle or you can do it
not
subtle in this case I ask it use the
power of three words and also one of
these three words should be a cini
principle reciate reciprocity in this
case so when I have fast free
fresh the word free fre is the from
Resto effect word you will you will see
free free is also C principle so in this
case I use the power of three words all
starting with an
F from west of effect and a c principle
now in my book I wrote I have an example
of this is a uh
solid exclusive stylish
chair now when you use exclusive for a
high price product the word exclusive is
what you read when I use this is a solid
safe stylish chair and you sell
children's chairs the word safe is
something which you want to read as a
parent again that's a from of effect and
you can use this use the power of three
words in a row a from of Effect one of C
line principles and instead of the word
free give me 10 different
words when you ask for 10 different
words there's always a word in there
which will fit in if it doesn't ask for
10 more for 10 more this will make it
very simple to do subtle psychology
using these different techniques which
have again never been combined before
and that's the the
strengths of generative AI I mean I
really like it
I'm I can't get enough of
it and practical examples of what what
I've been doing with this is uh creating
friend voice style
profiles and I use up to 80 elements so
most people think about tone of
voice but there's also tone there's
voice there's style there's language use
and these are just five elements which I
use when I create brand voice and style
profiles I also add marketing to that
but what I've been doing the last nine
months is creating
profiles where I can control the output
of any text so I can create a
journalistic profile and can create a
apple like profile uh I write like Ikea
add marketing to it I can use famous
bloggers uh let them sell products by
adding marketing to it again I'm using
what you saw in the previous spreadsheet
cini for caran fogler scarf model and
reactants uh this has taken me a lot of
time but I'm now able to control almost
any text the way I
want uh 80 elements is a
lot uh but this is really how I can
control the AI when people say right as
a copywriter or a
marketeer I think it's nice but it's
beginner because you don't know what
you're doing what I'm doing with these
profiles the ad elements is really
controlling how should this copyright or
right what language what style what tone
of voice plus up to 80 elements
including marketing so I'm not saying
it's a marketeer really specifying what
it should do it's me making the template
what it should do and this is really how
we should start using AI is controlling
everything uh but it's difficult the
deeper you go the more difficult it gets
and as I said this has taken me hundreds
of hours uh of time to understand this
uh lots of
frustration uh but it's it's working now
and uh it's a very nice example of
generative Ai and I was only able to do
this by asking the
model but we have Evolution and
evolution is very nice because uh once
you've mastered your prompting and you
think you understand how the model
interprets your
language we get the next model so we
have Chad GPT 3 and a half we have four
we have 4.0 five is around the corner we
have Claude we have
Gemini and it evolves and it means that
the words you were using half a year ago
the
prompts they might not work on the new
model or they might not do what you have
been doing and that's what I've seen in
Chad
GPT it gets better when using simple
prompting but again this gives the
average
results and to get out of that average
result I have to prompt even deeper I
have to tell it what it can do and what
it can't do and that's sometimes very
frustrating because you had a prompt
which worked very well but the more the
model develops the more it will make
assumptions based on what the average
user
wants and this means that probably next
year it will do a very good job for the
average
user but for the professional user who
wants to get out of that average result
you really have to understand what it
does using that prompt and then you have
to start negative prompting you can't do
this you can't do that you have to do
you have to do it this way you have to
do it that way and that means you have
to keep on prompting keep on asking what
it will
do and this will take a very long time
and as AI
evolves it means you have to go deeper
deeper and deeper into the model uh to
control it
and yes that you need creative thinking
for that logical thinking analytical
thinking and again lots of time now some
of these things can be automated because
we see agents now and I like the idea of
agents and you have agents talking to
each other I see people using uh AI to
make their
prompting again it's nice you can have
ai do your prompting but the AI can only
prompt it knows in its data set in the
previous
examples previous slides I gave you an
example of things I combined the AI
could never combine those because it
doesn't know how to combine them so yes
what I'm doing now
is I make my own
prompts combining things which have
never been combined before and then put
them in the AI to optimize my prompting
but yes I have the AI help me but I
still
first I make my own prompts and then
have ai optimize it and that's how it
will work for a long time if you want to
get away from the average
results so I hope you liked my
presentation if you have any questions
you can follow me on
LinkedIn send me an email and thank you
for your time and Happy prompting
everyone
Ver Más Videos Relacionados
AI in a Minute: Prompt Engineering
Master Prompt Engineering (Full Guide)
Generative AI vs. Conventional AI: Introduction For Operational Risk Professionals
Jascha Goltermann: The Impact of AI on UX Design - Hatch Conference 2023
The implications of AI on a Center of Excellence
The 9 AI Skills You Need NOW to Stay Ahead of 97% of People
5.0 / 5 (0 votes)