Skip to content
Snippets Groups Projects
Commit 47eb2f88 authored by marc Gay-Balmaz's avatar marc Gay-Balmaz
Browse files

Add notes

parent a87dcb96
Branches
No related tags found
No related merge requests found
......@@ -27,6 +27,13 @@
\end{multicols}
\end{frame}
\note{Who we are:
\begin{itemize}
\item Me
\item Orestis not there but prepared practical example, I'll present it today
\end{itemize}
}
\begin{frame}[fragile]
\frametitle{Generative AI - definition}
\begin{itemize}
......@@ -35,6 +42,13 @@
\end{itemize}
\end{frame}
\note{What means words I'll talk about
\begin{itemize}
\item LLM = Examples : chatGPT, Gemini, CoPilot,
\item Generative => G in chatGpt
\end{itemize}
}
\begin{frame}[fragile]
\frametitle{How does it work ?}
Example
......@@ -49,6 +63,15 @@
\end{itemize}
\end{frame}
\note{How it works
\begin{itemize}
\item Splits the input
\item labels kind of words
\item Compute probability of the kind of word
\item Compute probability of the next word
\end{itemize}
}
\subsubsection{Example}
\begin{frame}[fragile]
\frametitle{Example - Write me bibliography for the paper I juste wrote on AI in research}
......@@ -81,6 +104,15 @@
\item Prone to bias
\end{itemize}
\note{Details
\begin{itemize}
\item Definition byzantine - agent that may be malevolent or not, anyway it may poison the data, example add doubt in climate change
\item Concept = add \& mul, what it finds => hard to make sense of function found
\item Number of parameters = complexity, number of "information" stored in the NN
\end{itemize}
}
% \vspace{10pt}
\pause
......@@ -111,4 +143,4 @@
% }
%} % <-- end of the environment
%\end{lstlisting}
\end{frame}
\end{frame}
\ No newline at end of file
......@@ -11,6 +11,15 @@
\end{itemize}
\end{frame}
\note{Details
\begin{itemize}
\item Definition byzantine - agent that may be malevolent or not, anyway it may poison the data, example add doubt in climate change
\item Concept = add \& mul, what it finds => hard to make sense of function found
\item Number of parameters = complexity, number of "information" stored in the NN
\end{itemize}
}
\begin{frame}[fragile]
\frametitle{Good use cases}
\begin{multicols}{2}
......@@ -20,6 +29,15 @@
\item Easily formulate with the appropriate tone
\item Versatile
\end{itemize}
\note{
\begin{itemize}
\item Write as author, role or so
\item Correct orthograph
\item Summarize but may lose information
\end{itemize}
}
\pause
\vspace{10pt}
......
......@@ -3,13 +3,26 @@
\subsection{Limitations}
\begin{frame}[fragile]
\frametitle{Content quality}
\begin{multicols}{2}
\begin{itemize}
\item The model is not built to validate the content of the output
\item The user must check if the output is valid
\item $=>$ The user must be an expert on what they ask the LLM about% Fleche pour eviter trop parler
\end{itemize}
\framebreak
\includegraphics[width=7cm]{no_logic}
\end{multicols}
\end{frame}
\note{Talk about
\begin{itemize}
\item Study shows 39\% of errors missed by experts (last link in references)
\end{itemize}
}
\begin{frame}[fragile]
\frametitle{No logic}
\begin{multicols}{2}
......@@ -21,7 +34,7 @@
\framebreak
\includegraphics[width=7cm]{no_logic}
\includegraphics[width=7cm]{how_ai_thinks}
\end{multicols}
......@@ -36,6 +49,13 @@
\end{itemize}
\end{frame}
\note{Talk about
\begin{itemize}
\item Other study shows 62\% answers contain security breaches
\item These are critical, imagine what it means in your field
\end{itemize}
}
\begin{frame}[fragile]
\frametitle{Training}
\begin{multicols}{2}
......
......@@ -4,7 +4,7 @@
\subsection{A real use case}
\begin{frame}[fragile]
\frametitle{Let's write a paper with Copilot}
\begin{block}{What bing can help us with (if we ask it)}
\begin{block}{What Copilot can help us with (if we ask it)}
\begin{itemize}
\item<1,7> Choose a research topic
\item<2,7> Give ideas
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment