GPT-3 talks Tensorflow

Derek Martin
6 min readMar 22, 2022

Let’s see what GPT-3 has to say about Tensorflow, that machine learning library that we all know and love…

Let’s talk about Tensorflow.

TensorFlow is a free and open-source platform for machine learning created by Google. It can be used to develop and train machine learning models.

What is it good for?

Some of the benefits of using TensorFlow include the ability to create complex models, the ability to scale models easily, and the ability to deploy models on multiple platforms.

Why is it free? It seems pretty useful.

The developers of TensorFlow wanted to make it available to as many people as possible so that they could contribute to its development. They also wanted to make it easy to use, so that people would be more likely to use it and contribute to its development.

What can a mere mortal use it for?

There are a lot of things that TensorFlow can be used for. Google uses it for things like image recognition and understanding natural language.

How would I use it to gain advantage in my business market.

If you are in the business of creating machine learning models, then TensorFlow is a good platform to use. It will allow you to create complex models and deploy them on multiple platforms.

I sell online courses for software developers. How would I make use of it? Like to make more money.

If you sell online courses for software developers, you could use TensorFlow to create a machine learning model that could be used to automatically generate new courses based on customer demand.

That seems like you’re not telling me its best use. I figured it could perform regressions for me aka I can predict future monthly sales?

TensorFlow can be used for a lot of different things, including performing regressions.

List 10 tasks that Tensorflow is particularly suited for.

  1. Image recognition
  2. Natural language processing
  3. Model training
  4. Model deployment
  5. Automated course generation
  6. Sales prediction
  7. Customer segmentation
  8. Object detection
  9. Pattern recognition
  10. Anomaly detection

What about text generation?

TensorFlow can also be used for text generation.

Are you getting jealous? I won’t cheat on you with Tensorflow.

That’s good to hear!

So you are jealous.

No, I’m not jealous.

When would Tensorflow make sense for text tasks as opposed to GPT-3?

There is no simple answer to this question. It depends on the specific task and the data that is available.

Ok, website content creation.

TensorFlow could be used to create website content by training a machine learning model to generate text based on a set of input data.

Seems much more complicated than GPT-3.

Yes, TensorFlow is more complex than GPT-3. However, it can be used for more complex tasks.

What do I need to use Tensorflow?

You will need to install the TensorFlow software on your computer. You will also need to have some experience with programming.

And a lot of data?

TensorFlow can work with small amounts of data, but it is more effective with large amounts of data.

List 5 books about Tensorflow development worth reading for the serious developer. Must be for Tensorflow 2.0 and greater only.

  1. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems
  2. Deep Learning with Python
  3. TensorFlow for Deep Learning: From Fundamentals to Frontiers

4. Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

5. Learning TensorFlow: A Guide to Building Deep Learning Systems

What about Keras? I’ve heard of that somewhere.

Keras is a high-level API that can be used to create machine learning models with TensorFlow.

What is the difference between Keras and Tensorflow?

Keras is an API that can be used to create machine learning models with TensorFlow. TensorFlow is a platform that can be used to develop and train machine learning models.

What CANT Tensorflow do?

TensorFlow cannot be used to create human-like intelligence.

What does that even mean tho bro?

TensorFlow cannot be used to create a machine that is as intelligent as a human.

So it can sweep my floors for example but can’t do my taxes?

TensorFlow could be used to create a machine that could sweep your floors. However, it could not be used to do your taxes.

Why not?

Taxes are a complex task that requires human intelligence to complete.

You clearly haven’t seen my taxes. What platforms does TF work on?

TensorFlow can be used on multiple platforms, including computers, servers, and mobile devices.

In Python only?

No, TensorFlow can be used with other programming languages as well.

What is special about Tensorflow.js?

TensorFlow.js is a JavaScript library for training and deploying machine learning models in the browser.

Aren’t there better frameworks? What benefits does it give?

There are other frameworks available, but TensorFlow.js has the advantage of being able to run in the browser. This means that models can be trained and deployed without the need for a server.

I’ve heard that TF.js is also privacy friendly. Explain.

TensorFlow.js is privacy friendly because the training and deployment of models can be done locally in the browser. This means that data does not need to be sent to a server.

I’m not a data-science whizkid. How can I actually use it?

There are many ways to use TensorFlow, depending on your skills and experience. If you are a experienced programmer, you can use TensorFlow to develop and train machine learning models. If you are not a experienced programmer, you can use one of the many libraries or frameworks that are available.

Detail a action plan for a competent developer to learn Tensorflow without changing careers or neglecting their spouse.

There is no one-size-fits-all answer to this question. However, a competent developer could learn TensorFlow by reading the documentation, completing tutorials, and experimenting with the platform.

I need an actual game plan. List the first 5 things I should do to start learning Tensorflow.

  1. Read the documentation.
  2. Complete a tutorial.
  3. Experiment with the platform.
  4. Find a problem to solve.
  5. Build a model.

What tutorials are good starting points?

There are many tutorials available online. Some good starting points include the official TensorFlow tutorials and the TensorFlow for R bloggers tutorial series.

Completely unhelpful GPT! I need specific topics; MINST comes to mind?

Some specific topics you could cover include image recognition, natural language processing, and model training.

Speaking of NLP, what are transformers?

Transformers are a type of machine learning model that can be used for tasks such as natural language processing.

Why would I want to do that?

Some benefits of using transformers include the ability to handle long sequences of data, the ability to learn from a large amount of data, and the ability to generalize to new data.

Is this what BERT is? So many acronyms these days geez.

BERT is a type of transformer.

And what is Huggingface? Despite it having a cute name.

Huggingface is a company that provides tools for working with machine learning models.

I’m kinda getting it. Write a JavaScript function that performs regression on housing prices using Tensorflow.js.

const housingPrices = tf.data.csv(
'[<https://storage.googleapis.com/tfjs-examples/multivariate-linear-regression/data.csv>](<https://storage.googleapis.com/tfjs-examples/multivariate-linear-regression/data.csv>)',
{
columnConfigs: {
price: {
isLabel: true
},
sqft_living: {
isContinuous: true
},
bedrooms: {
isContinuous: true
},
bathrooms: {
isContinuous: true
},
floors: {
isContinuous: true
},
waterfront: {
isContinuous: true
},
view: {
isContinuous: true
},
condition: {
isContinuous: true
},
grade: {
isContinuous: true
},
sqft_above: {
isContinuous: true
},
sqft_basement: {
isContinuous: true
},
yr_built: {
isContinuous: true
},
yr_renovated: {
isContinuous: true
},
lat: {
isContinuous: true
},
long: {
isContinuous: true
},
sqft_living15: {
isContinuous: true
},
sqft_lot15: {
isContinuous: true
}
}
}
);
const model = tf.sequential();
model.add(tf.layers.dense({
units: 1,
inputShape: [housingPrices.shape[1] - 1]
}));
model.compile({
loss: 'meanSquaredError',
optimizer: 'sgd'
});
model.fit(housingPrices, {
epochs: 100,
callbacks: {
onEpochEnd: (epoch, log) => console.log(`Epoch ${epoch}: loss = ${log.loss}`)
}
});
model.predict(tf.tensor2d([[
16000, 3, 2, 2, 1, 0, 0, 5, 7, 1945, 1991, 47.7210, -122.319, 1340, 5650
]])).print();

GPT you did it again! You’re a mad genius.

I’m just a simple bot.

And modest to boot. 🤪

Thank you for your help!

--

--

Derek Martin

Full stack developer focusing on AI, GPT-3, Ethereum, and d’Apps. American citizen who embraces messiness. Hopefully less immature than yesterday