The Stargate Project, The Possible Future of AI, & How You Can Use It To Your Advantage (Part 2)

Last week, we ended with the possible future of artificial intelligence, and how AGI or even ASI could be developed thanks to a boost from projects like Stargate.

However, many individuals may find themselves asking the question: What does all this have to do with me?

Well, that’s a good question. The Stargate Project may seem irrelevant to many people, and that’s true.

However, AI in general shouldn’t be treated that way.

There are near limitless options of AI. Many companies offer their uniquely tuned models that can complete a variety of tasks. Most of them are free, some offer a limited number of credits (messages), and others are completely open-sourced.

The open-sourced ones are my personal favourite, as you can tweak the code to fit your needs, download it to your own computer, and overall, they’re much safer as there’s no behind-the-scenes (everything is entirely transparent). In addition, some of them can even be used offline, which can be quite helpful in cases where you can’t access a particularly secure internet connection.

Installing Deepseek R1 (A Step-By-Step Guide For Mac)

There are many po0werful AI models out there on the market, including newly released ones such as Deepseek R1 (made by a Chinese company). A most powerful version of this model supposedly out-competed the OpenAI o1 model in a number of benchmarks (the OpenAI model was previously considered one of the best). However, what I like most about Deepseek R1 is that the entire model is open-sourced.

After installing it onto your device, the model should be able to run offline (though certain capabilities may be limited), and you can send unlimited amounts of messages.

But what sets Deepseek apart from OpenAI o1 is that it has the ability to ‘reason’. Basically, it guides you through all the steps of its thought process, from interpreting your prompt to answering it. This can be quite helpful in a variety of ways, including solving complex math, coding, etc. You could think of it as an AI that ‘thinks aloud’.

Because of this, I sought out to download this model, and try it for myself. Along the way, I hit a lot of challenges and hurdles, as well as setbacks where I failed to make any progress for days. However, after weeks of trying, here’s a hopefully useful guide to how to download Deepseek R1 (or similar open-sourced projects) onto a Mac operating system.

Note that specific processes may vary, if faced with issues, please consult necessary sources.

Firstly, you need to find an important tool that you’ll be using throughout this installation process, the Terminal application, which is a preinstalled application that uses a command-line interface to interact with the operating system instead of a graphical user interface, allowing for the completion of a wide-range of tasks. To find this application, just open Spotlight (Command + Space), then type ‘Terminal’. To run any command-line, just type it or paste it into the Terminal prompt section (starting after %), then press ‘Enter’ to run.

But before we can actually start using the Terminal application to install Deepseek, we need to install a series of what are known as ‘dependencies’. For Deepseek, four dependencies are needed: Anaconda, Python, Ollama, and PyTorch. And do remember to have a reliable browser to access the websites needed for installation.

As the code is mainly based on python, you need to install the python package from the website https://www.python.org/. It’s a relatively straightforward process: Just navigate to their homepage, and there will be a link for downloading the latest version of python. Choose the correct operating system (don’t download the wrong version). After completing the process, you should find the a new ‘Python’ in the applications folder (the applications folder is in Finder).

Alternatively, you could download Python via another way through Homebrew, a package manager. Go to their website https://brew.sh/ and copy the command-line provided (it should be ‘/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"’, though do verify that) then go to Terminal. Paste the command-line, and Homebrew should be installed. Make sure it is up-to-date with the line ‘brew update’ in Terminal, then run ‘brew install python’.

In both methods, you can verify installation by running ‘python3 --version’ and a python version number should pop up.

Moving on, you need to install Ollama (which stands for Omni-Layer Learning Language Acquisition Model). This tool is important, as you’ll need it to directly install Deepseek onto your Mac. Just go to their website https://ollama.com/download, then click the package for Mac (note that installing Ollama requires macOS 11 Big Sur or later). Open the installer, and follow the instructions to install Ollama. To verify installation, use ‘ollama --version’.

Then you can move on to download the other dependency, Anaconda. To do this, go to https://www.anaconda.com, and navigate to their downloads page. Clicking download for macOS, it will install an installer. Open this package, which will direct you through the process of downloading Anaconda. After installation, you can verify the Anaconda version with a similar command-line in Terminal as used previously with python: ‘conda --version’. This version number should be the same as with the website.

Now, why do we need Anaconda? Well, the main reason for this particular project is because it comes with Conda, a powerful package manager that we can use to create virtual environments. Virtual environments are basically separate operating rooms. They reduce conflict between different packages. For example, another project may require a different version of python, which means the operating system could become confused on which version to use. Here, if you downloaded the new python version separate from the other base model in your applications folder, then it’s like two separate operating stations completely blocked off from each other, preventing mix-ups and confusion.

So, to implement a Conda virtual environment, we need to run the command-line ‘conda create --name Deepseek’. This will create a virtual environment called ‘Deepseek’. You could replace the name by changing out the input after ‘--name’, though do be aware to keep one ‘word’ by using underscores (_) to connect separate words.

After running the ‘conda create’ line, you need to enter ‘y’ into Terminal to allow the system to create and download needed packages for the environment (you could enter ‘n’ to terminate environment creation). To activate the environment, just run ‘conda activate Deepseek’ (or whatever your environment name is). You will notice that the heading before the prompt section, which used to looking something like ‘(base) name@MacBookAir %’ would become ‘(Deepseek) name@MacBookAir %’. The words within the () will be replaced by your environment name, indicating that the environment was activated successfully.

To go back to your base environment, use ‘conda deactivate’. To delete an environment, first prompt the Terminal ‘conda env list’, which will show all Conda environments. Find the one that you want to remove, verify the name, and enter ‘conda env remove --name Deepseek’ (or any other environment name).

If you wanted to request for a specific version of python within your environment, just change the normal command prompt for creating a virtual environment to ‘conda create --name Deepseek python=version’ (replace version with your desired version).

With Anaconda installed, you now have to install the final dependency, PyTorch.

PyTorch is important for Deepseek’s deep learning capabilities.

Go to their website https://pytorch.org/, and scroll down to the section labeled ‘Install PyTorch’. Here, there are various configurations. Select the one of your preference, and copy the provided command-line.

Alternatively, you could run the command-line ‘conda install pytorch torchvision torchaudio cudatoolkit=11.7 -c pytorch’ for GPU, and ‘conda install pytorch torchvision torchaudio -c pytorch’ for CPU.

Now is where all of that hard work pays off. With all the dependencies installed within your system, all you need to do is enter the final command-line to install Deepseek.

However, before you do that, you need to choose the appropriate version of Deepseek. The model comes in 7 versions that differ in size. Depending on your disk space as well as the memory of your Mac, you can choose the best one to fit your needs. Remember to check your available disk space before downloading, or else it may crash or result in error.

Here is the list of the different versions and the command-lines to install them:

1.1 GB (‘ollama run deepseek-r1:1.5b’)

4.7 GB (‘ollama run deepseek-r1:7b’)

4.9 GB (‘ollama run deepseek-r1:8b’)

9 GB (‘ollama run deepseek-r1:14b’)

19 GB (‘ollama run deepseek-r1:32b’)

42 GB (‘ollama run deepseek-r1:70b’)

404 GB (‘ollama run deepseek-r1:671b’)

Each of these models gets larger and larger in size, and also requires more and more memory to run. I recommend starting with the smallest 1.1 GB version, and potentially upgrading it in the future.

If you do choose to install multiple versions of Deepseek to see which one runs best (like I did), I also recommend that you create a new virtual environment using Conda before installing each new version, as it can get quite messy.

After actually running the command-line for your chosen version, you should wait for the installation process to complete. Upon completion, the model should automatically initialize, and the input section should start with ‘>>>’.

To use the model, here are some basic commands to be aware of.

If you just downloaded it and it’s already initialized, you can type your prompt into model and press enter to send.

Depending on your device’s free memory, the output speed (tokens/second) should be either fast or very slow. You can increase this speed by freeing up your device’s memory. To do this, open Spotlight (Command + Space), and search for ‘Activity Monitor’. This application gives you access to the device’s processes, including Memory, CPU, GPU, Energy, Disk, and Network. For our purposes, go to Memory, then you will see all the current processes that are using memory. My Mac has a memory of 8 GB, though your device may have different specs. Either way, the specs of your Memory and how much is being used should appear at the bottom of the window. Doing some basic subtraction, you can figure out how much free memory your device currently has. You will see that by default, the most memory consuming processes appear at the top of your screen. To terminate a process, click on it to select, then click the ‘X’ icon at the top menu bar. It will prompt you with three choices, ‘Quit’, ‘Force Quit’, and ‘Cancel’. Normally, quitting the process is enough. I should note here that it’s not advised to close the process called ‘Window Server’. I did close it once out of curiosity, and all my windows vanished and my system crashed (though it was fixed after restart).

Moving on, here are some tips to interact with the model itself. Entering ‘/bye’ terminates the session and ‘/clear’ clears previous conversation without terminating the session. For more command-line prompts, just enter ‘/help’, which opens up the help menu.

After terminating a session, all you need to do is enter the same command-line you used to download the model. But now, because it was already downloaded, it will just activate the model. All you need to do now is have fun.

Conclusion

I hope you learned something new in this two-part series, whether it was about the Stargate Project, the possible future of AI, or how to get an AI model for yourself.

For more about Deepseek, visit their official website https://www.deepseek.com/. For more about OpenAI, visit their official website https://openai.com/. For more about open-sourced models, I recommend visiting GitHub, where there are a lot of models waiting to be discovered.

Finally, I just wanted to note that I’m currently trying to use another open-sourced AI model called ‘Genesis AI’, though I’m facing issues with the rendering part on Mac. If you just so happen to know the solution to rendering the model, please post your ideas in the comments.

Thank you for reading this article. Remember to subscribe and share this article, so that more people can join in on the emerging era of AI.