Generative AI is just the Beginning AI Agents are what Comes next | Daoud Abdel Hadi | TEDxPSUT
Summary
TLDRこのスクリプトは、AIと特に機械学習の進歩を説明し、GPT-3の登場以来の大きな一歩を強調しています。AIは従来のように特殊なタスクに限定されていなく、自然言語を使用してパターンを認識し、理論的に人間のようになります。AIの現在の状態と、将来的には自動化されたエージェントとして機能する可能性についても説明しています。これらのエージェントは、エンドツーエンドのワークフローを自動化し、人間の介入を最小限に抑えます。スクリプトでは、AIが私たちの生活やビジネスにどのように影響を与えるかを具体的に示しています。
Takeaways
- 🎓 脚本の主人公はAIの硕士号を接近し、AIが真の知能を持ち得るまでの距離が遠いと感じていた。
- 🤖 AIは診断、詐欺検知、交通最適化など多くの分野で役立つが、特定のタスクに特化しており、人間のように一般化能力が低い。
- 🚀 2年後にOpenAIがGPT-3をリリースし、AIの大きな進化をもたらした。
- 📚 GPT-3は書籍、記事、研究論文などをデータとして、強力なコンピュータで訓練された大規模な言語モデル。
- 💡 GPT-3は自然言語で文章を書く、質問に答え、コードを読み書きすることができ、知能の兆候を見せる。
- 🤔 GPT-3は完璧ではない。事実をでっち上げる「幻觉」を起こすことがあり、情報が古く、基本的な数学の問題にも苦戦する。
- 🧠 人間は問題解決能力を持っている。それは知識だけではなく、計画能力、問題を小分けする能力、行動の結果を反映する能力、ツールの使用能力が関与している。
- 🤖 AIをチャットボットではなく、自主的なエージェントとして考えると、ワークフローを自動化し、人間の介入が少なくなります。
- 🛠️ エージェントは、AIが使用するツールを自分で選択し、使い方を決定し、自主的に実行する。
- 🌐 エージェントの可能性は、ウェブ開発やデータ分析、旅行計画など、様々なタスクを自動化することにある。
- 📈 技術の発展により、エージェントは私たち生活にますます近づいている。MicrosoftのCopilotやShopifyのSidekickなどが既に存在している。
- 🌟 AIの発展により、コンピュータとの交渉方法が変わり、より協力的な関係になる可能性がある。
Q & A
AIが真の知能を持ち得るに至ったカギは何ですか?
-AIが真の知能を持ち得るカギは、大規模な言語モデルGPT-3の登場です。これにより、AIは自然言語を理解し、様々なトピックの質問に答え、コードを書く、または記事、歌、詩を書くことができるようになりました。
GPT-3がリリースされる前に、AIが人間の仕事を完全に自動化するという考えはどのように見えましたか?
-GPT-3がリリースされる前は、AIが人間の仕事を完全に自動化するという考えは非常に遠いものと見えました。AIは特定のタスクに特化したものであり、人間のように他のタスクに簡単に一般化することができませんでした。
GPT-3がもたらしたAIの進化は何ですか?
-GPT-3がもたらしたAIの進化は、AIが単なる専門家から、自然言語を用いて様々なタスクを理解し、実行するようになることでした。これにより、AIはプログラミングする必要なく、自然に文章を書き、問題を解決する能力を獲得しました。
GPT-3が得意とすることは何ですか?
-GPT-3は自然に文章を書き、幅広いトピックの質問に答え、コードを書くことができるようになっています。また、記事、歌、詩など様々な形の書き込みを行うことができます。
GPT-3がまだ完璧ではない理由は何ですか?
-GPT-3はまだ完璧ではない理由は、事実を架空することや、情報が古くなること、基本的な数学問題を扱うことが難しいこと、またはマルチタスクでの課題に対応するのが難しいことが挙げられます。
AIが人類と同様の問題を解決する能力とは何ですか?
-AIが人類と同様の問題を解決する能力とは、将来を予測し、問題を小さく分け、行動の結果を反省する能力、そしてツールを使用して目標を達成する能力を指します。
AIエージェントが何を自動化するのですか?
-AIエージェントは、ワークフローを自動化し、人間の介入を最小限に抑えます。これには、計画の立案、行動の実行、結果の反映、そしてツールの使用が含まれます。
AIエージェントが使用するツールは何ですか?
-AIエージェントは、人類が使用する同様のツールを使用します。これには、ウェブブラウザ、エクセル、プログラミングツール、データ分析ツールなどが含まれます。ただし、AIはこれらのツールを自動的に選択し、使用する方法を決定します。
AIエージェントがビジネスにおいてどのように役立つか?
-AIエージェントは、ビジネスにおけるデータ分析、ウェブサイトの構築、旅行の計画、エッセイの執筆など、様々なタスクを自動化することができます。これにより、ビジネスはより効率的になり、コスト削減や新しいサービスの創出が可能になります。
AIエージェントの普及がもたらす可能性は何ですか?
-AIエージェントの普及は、コンピューターとのインタラクションの進化をもたらす可能性があります。これには、グラフィカルインターフェースからAIアシステッドインターフェースへの移行や、より多くの人々が革新を起こしやすくなることなどが含まれます。
AIの発展がもたらす可能性についての個人的な考えは何ですか?
-AIの発展は、技術スキルが民主化され、革新の壁が低くなることを意味しています。これにより、より多くの人々が製品やサービスを作り、問題を解決することができます。AIは私たちよりもツールを使用するのに優れていますが、それは私たちがより大きなpicturedに焦点を当て、本当に重要なツールを使用する機会を与えるものです。
Outlines
🤖 AIと機械学習の進化と課題
この段落では、著者は自分の人工智能(AI)の修士課程を終えた当時、AIと機械学習が様々な分野で役立っている一方で、それらは特定のタスクに特化しており、人間のように一般化することができないという問題点に触れています。AIが人間の仕事を完全に自動化するというアイデアは信じがたいものだと感じていたが、OpenAIがGPT-3という大規模な言語モデルをリリースしたことで、その考え方が劇的に変わりました。GPT-3は自然言語を用いて文章を書く、質問に答え、コードを読み書きすることができる能力を示しており、知能の兆候を示しています。
🤖 代理人(Agents)の概念とその可能性
この段落では、AIをツールとして使用することで、代理人(Agents)がどのようにしてタスクを自動化するかについて説明されています。代理人は、人間がAIにタスクと最終目標を説明し、AIが必要なツールを選択して使用する方法を計画し、自主的に実行します。これにより、Web開発者を雇うのではなく、AIがビジネスを説明してウェブサイトを構築するなど、より迅速で効率的な作業が可能になります。また、データ分析や旅行計画など、様々なタスクを自動化できる可能性が示されています。
🤖 既存のAI代理人の例とその将来
最後の段落では、既に存在するAI代理人の例が挙げられています。MicrosoftのCo-PilotやShopifyのSidekick、Hyperwriteなどが、それぞれのアプリケーションを通じてAIを活用してタスクを自動化する例です。また、GPT-3自体も様々なエージェントを提供しています。これらの代理人がより普及し、より高度なAIとして進化すると、コンピュータとの相互作用の考え方乃至びに社会全体が変わることになる可能性についても触れられています。最終的には、AIとの協働関係が重要であり、人間の創造性や経験を活かすことが強調されています。
Mindmap
Keywords
💡AI
💡machine learning
💡GPT-3
💡intelligent agents
💡natural language processing
💡code
💡multitasking
💡digital labor
💡innovation
💡collaborative
Highlights
Six years ago, the speaker was completing their master's degree in AI, feeling that true intelligence with computers was far away.
AI and machine learning have been revolutionary in diagnosing illnesses, detecting fraud, and optimizing traffic flow.
AI has traditionally been a specialist in specific tasks, not generalizing well to others, similar to humans.
OpenAI's release of GPT-3 marked a massive leap forward in AI, demonstrating more generalized intelligence.
GPT-3 can write naturally, answer questions on various topics, and even read and write code, showcasing impressive capabilities.
Despite its abilities, GPT-3 and similar AI are not perfect, can make mistakes, and struggle with basic math and multitasking.
The speaker suggests that intelligence is not just knowledge but also the ability to plan, reflect, and use tools effectively.
AI can be thought of as autonomous agents, designed to automate workflows with minimal human intervention, similar to how humans approach problems.
Agents use language models to plan tasks, reflect on outcomes, and use tools, much like humans do in practical operations.
The potential of agents includes building websites, making business decisions, and planning trips, all without human knowledge of specific tools.
Agents are digital labor, capable of browsing the web, navigating files, using applications, and controlling devices on our behalf.
Everything on our screens is formed of code, allowing AI to use and combine various tools and applications creatively.
Agents operate through a cycle of planning and executing actions using language models, breaking down tasks into a series of actions.
Examples of agents include Microsoft's Copilot, Shopify's Sidekick, and Hyperwrite, each automating specific tasks in their respective domains.
As agents become more widespread, intelligent, and sophisticated, they may change how we think about computers and human-computer interaction.
The democratization of skills through AI empowers more people to innovate and build solutions, lowering barriers for individuals and small businesses.
AI's ability to use tools quickly and efficiently may lead to a collaborative relationship with humans, focusing on bigger-picture thinking and creativity.
Transcripts
[Music]
about six years
ago I was approaching the end of my
master's degree in Ai and despite
studying a variety of different projects
involving machine learning genetic
algorithms and even generative AI
I couldn't help but feel like we were
still so far away from creating true
intelligence with
computers yes Ai and more specifically
machine learning has been a revelation
in so many ways helping us diagnose
illnesses detect fraudulent activity
optimize traffic flow and so much more
but if you've worked with AI in practice
you'll know AI has always been more like
a specialist very good but at very
specific tasks and unlike humans they
don't generalize very well to other
tasks so to me at the time the idea that
AI would completely automate people's
work seemed quite far-fetched
except I was
wrong very wrong only two years later we
saw a massive Leap Forward for AI That's
when open AI introduced the world to
large language models with the release
of gpt3 the predecessor to chat
GPT it was basically one big
experiment what would happen if we
gathered all the data we can find every
book every article every research paper
and trained in AI with the most powerful
computers
available many of you will have tried
chat jpt and seen for yourselves what we
get are signs of intelligence even
without programming it to do so
it can write in a natural way it can
answer questions on a huge range of
topics it can read and write code and it
can do different forms of writing like
articles songs and
poems what's impressive is it does those
things surprisingly well but what's
arguably more impressive is the way it
can reason and recognize patterns in
similar ways to
us using just natural language we can
ask a question or give it an instruction
and somehow it understands our
request AI now is no longer just a
specialist and that in
itself is a huge
milestone now many of you will have
tried Chachi PT or something similar and
thought this is great for brainstorming
ideas writing content doing some editing
and maybe even answering some queries
but that's just the tip of the iceberg
of what generative AI can actually do
to understand just how far this
technology can go I'd actually like to
start with what it doesn't do very well
first and you'll see why in a second now
a lot has been said about whether this
technology is actually
intelligent that's because they're not
perfect they can make
mistakes for example it can completely
make up facts also known as
hallucinating the information it has
isn't always up to date
and surprisingly it can struggle with
even basic maths and it can often times
struggle with
multitasking these can be big issues
depending on what you're trying to do
but if we're going to be
fair we're not perfect either many of us
can struggle with those same things as
well but maybe a counterargument to that
is we still manage to solve problems and
get things done why is
that
that's because our intelligence isn't
confined to just our knowledge and what
we know it encompasses a lot more like
our ability to plan ahead and break down
problems into smaller
problems it's down to our ability to
reflect on the outcomes of our actions
and of course down to our ability to use
tools to help us complete a given
goal now here's where things get
interesting what if we try try to use or
replicate the way humans approach
problems using language models it's at
this point when we stop thinking about
AI like chat GPT as just a chatbot that
requires constant human input to
complete a task and instead think think
of them as agents or autonomous
agents agents are designed to automate
workflows end to endend with little to
no human intervention and they do so
by planning their tasks reflecting on
the outcomes of their actions and using
tools to help them out so very similar
to us to understand what agents actually
do let's think about how we generally
operate practically everything we do
nowadays involves our phones and our
computers the applications and programs
on our devices act as tools to do
certain
things so if I'm writing an essay I
might use a web browser to access the
intern internet to research a topic if
I'm an
accountant I could use a different set
of tools maybe Excel or an accounting
software if I'm a programmer again I
have a different set of tools but we're
basically constantly using a variety of
different tools to help us with a given
task and this is where agents are a bit
different instead of us using those
tools we just describe to an AI what the
task is and what the end goal is and
then then it plans which tools it needs
to use and how to use them and then it
actually does it on its own not only can
they complete the task much quicker than
we can but in theory we wouldn't even
need to know how to use these tools in
the first place so imagine this you're
an entrepreneur you've just started a
business and you need a website but you
have no idea how to build one but
instead of hiring a web developer you
just describe your business to an AI you
describe how you want the website to be
and then the AI uses the same tools the
web developer does and builds the entire
thing in a matter of
seconds or maybe you've been collecting
a bunch of data and you want to make
Better Business decisions instead of
hiring a data analyst the AI is using
data analysis tools to answer your
business queries instantly or finally
maybe you're just looking to travel
somewhere nice so you ask an AI to plan
your trip for you finally you flight
options accommodation and even plans
your activities once you're there that's
the potential of agents and believe it
or not we're not far off from that
world agents are like digital labor
capable of automatically browsing the
web navigating our files using our
applications and potentially even
controlling our devices for
us so how is this all possible it sounds
a bit like s science fiction but because
of today's technology it's simpler than
it
seems first an important thing to
understand is that everything we see on
our screens is just a visual
representation we can make sense of so
but in the background everything is
formed of code so for every button and
for every action you can do there's a
piece of code Associated to it and
here's just a couple of examples this is
the programming way of using chat GPT so
instead of going on the website and
using their interface you can just run
this line of code and it gives you gpt's
answer similarly this is how you might
do a Google search so uh just an
alternative to Google search and this is
how you might use or create a Word
document the only reason why I show this
is because when things are formed of
code it means we can assemble programs
that combine different different
functionalities and different
applications in creative and interesting
ways and and that's a core aspect of how
agents work by its ability to use
different tools
interchangeably and so if we look at the
framework of a of an agent it consists
of a set of actions it can do and again
these actions are just code just like we
saw in the last slide and the language
model like chat GPT and first the user
gives a request in this example it might
be book me the cheapest flight to
London and then we start a cycle of
questioning and answering with the
language model and we can ask a question
as simple as this based on the user's
request what do you think the next
action should be because language models
are so smart we can get them to break
down a task into a series of actions it
needs to perform to complete the end
goal and so in this example it might be
searching flights and again because
language models can read and WR code we
can get them to format the input just
how we need them to be to execute that
code with no errors then once it does we
can get the output which might be a list
of flights and their prices and then we
repeat this questioning process again
based on the information you have now
what do you think the next action should
be and then we repeat this process over
and over until the task is actually
complete so agents are just a feedback
loop of planning and executing actions
using language models and I mentioned
that we're not far off having agents but
in reality they're already
here for example Microsoft's co-pilot is
an example of an
agent within Excel you can just use
natural language to get it to analyze
your spreadsheets and reports without
needing to know all the fancy formulas
and all the complex
functionalities similarly Shopify has a
sidekick that helps you build a website
using
AI hyper write acts like a personal
assistant that can book flights for you
order takeaway and even organize your
emails for you and even chat GPT itself
has a whole catalog of Agents known as
gpts and this is just the start I
believe many more businesses will start
incorporating agents both internally and
as part of their products and services
language models are getting cheaper and
cheaper every year to the point where
they're virtually free and not only that
but they're also very accessible and
easy to use so practically anyone with
basic programming knowledge can get
started with building an
agent but as agents become more
widespread more intelligent and more
sophisticated it'll likely change the
way we think about computers in the
first
place in the same way that the
transition from a command line interface
to a graphical interface completely
revolutionized the way we interact with
computers perhaps the next evolution of
that is a kind of AI assisted interface
maybe like Tony Stark and his AI
Jarvis there's no doubt that a world
full of intelligent assistants and
agents is going to be
strange the technical skills we once
thought are unique to us are now being
outsourced to AI even as a data
scientist myself I wonder how long
before an AI can do the things that I
can do it is a scary thought but at the
same time as an optimist I can't help
but feel like it's also incredibly
empowering with our skills
democratized the barriers to Innovation
are lower than ever
more people can participate in creating
Solutions and building things that were
once only in the hands of large
corporations and specialized
professionals and in the same way that
Jarvis doesn't replace Tony Stark I
believe our relationship with AI will
always be one of kind of a collaborative
one sure AI might be better and quicker
at using the tools than us but that
gives us an opportunity to focus on the
bigger picture and use the tools that
actually matter our creativity Ingenuity
and Human
Experience thank
[Applause]
you
5.0 / 5 (0 votes)