Future of EBVM

Question 1

In 2016, there was a study that estimated there were around ______ _________ papers.

Answer 1

38 million

Question 2

As of 2016, how many of these papers ere cited? Who funded these papers that were published?

Answer 2

• ~50% have never been cited
• 46% of clinical trials are funded by pharmaceutical companies

Question 3

We are drowning in info

Question 4

Money ball the movie was a big deal
- About baseball and how it was revolutionized by big datasets.
- Picked based on family image, etc.
- One person came, used numbers to convert papers with baseball stats (very important stat: how many times does a player successfully get to first base?)
- Put together most unlikely team ever but had the longest streak of wins out of any other team.

Snowden data leak
- Troves of data released.

Panama documents: leaked to illegal funds and documents all over the world.

Wiki-leaks

Alexa –> Listen to us at home.

Question 5

Google glasses
- Worked but too ahead of its time

Augmented reality
- Put phone up to something and it pops up information.

Question 6

What does this image represent?

Answer 6

Big data

Question 7

What does this image represent?

Answer 7

big data
- associated with connectiveness to phone

Question 8

What do we mean by big data?

Answer 8

Data mining and knowledge discovery.
Data mining: extacting patterns adn trends from veyr large datasets

KD: looking for poential hypotheses and trends in big data

Stats: we are sampling from population, so need stats to see how well dataset represents the larger population.

Databases: collecting huge amounts of data. Any time you are entering anything digitally you can store it which is very powerful.

Visualization: box plots, graph things over time. looking for spacial and temporal trends, risk factors, correlations.

Info retrieval: making sense of databases.

Machine learning: you are allowing certain algorithms to find trends on its own. E.g. regression analysis - a type of machine learning b/c it is able to find predictions on historical empirical datasets; it is the prediction that explains the data the best.

AI: driven on a combo of machine learning in the past + going forward training a dataset and learning that some thigns are right and osme are wrong, so the machine can train itself again. The algoorhtim keeps repeating itself, adding moreinformation, and learning. It is called intelligent because it is training itself and changing algorithim.

Question 9

“The fundamental tools of AI shifted from logic to probability in the late 1980s, and fundamental progress in the theory of uncertain reasoning underlies many of the recent practical advances”
- Peter Norvig, 2011
(Director of Research at Google)

Answer 9

2011 - Shift from machine learning –> probabilities which is more along the lines of AI

Question 10

1997 – IBM’s ‘Deep Blue’ beat the world _____ champion
* Running every possible outcome within a well-______ set of _____.

Answer 10

chess, defined, rules

Question 11

2011 – IBM’s ‘WATSON’ beat the world ______ champions
* Used ________ language processing, and used every single document via _________ retrieval, and _________-learning technologies to _____-domain question answering

Answer 11

Jeopardy, Natural, information, machine, open

Question 12

How can IBM help oncologists?

Answer 12

Trained watson in oncology.
Analyze patient medical record to help clinician evidence based personalized treatment options.
- family history, notes from prior office visits, and test results
- highlights aspects of her notes that may be significant based on current research
- Clinician can see where this info is being pulled from.

Uses natural language to understand context in a file and make inferences based on attributes—> attributes.

Question 13

2016 – Google’s AlphaGo beat the world _____ champion
* More play possibilities than all the atoms in the universe
* Monte Carlo tree search and deep neural network technology (12 layers)

Answer 13

2016 – Google’s AlphaGo beat the world Go champion
* More play possibilities than all the atoms in the universe
* Monte Carlo (random component) tree search (different pathways you select) and deep neural network technology (has 12 layers that can talk to each other)

Question 14

2022 – ChatGPT (OpenAI)
* Chat Generative Pre-trained Transformer… an AI chatbot (version 3.5)

Answer 14

Uses 3 layers of neural networks

Question 15

What is a neural network?

Answer 15

Neurons are all connected to one another in a 3d matrix. Different parts of the brain light up when you think of something.
- One neuron can send different voltage signals to another neuron.
- We can remember many more things than neurons exist.
- The signal that the neuron gives to their neighbors that is more important.
-

Question 16

Neural network technology (3 layers)
* Step 1 = unsupervised learning (groups internet information – until 2021)
- usually with natural language interpretation you have no context. This has been a major setback. So if you can put in a context and cluster texts around one another you can apply context to things.
* Step 2 = supervised learning (100s people correcting answers)
* Step 3 = self-training and reinforcement when we get an output; say if something is right or wrong.

Question 17

Is chatgpt good at critically appraising a topic?

Answer 17

No! not at all