This, the quarter-thousandth edition of the Weekly Head Voices, covers two weeks from Monday November 21 to Sunday December 4, 2022.
I am happy that my ever-evolving life philosophy (as well as my approach to health and wellness) includes a growing section on the importance of hard and challenging work with great colleagues, because these past few weeks have really been delivering in that respect.
The rest of this post consists of two main parts: The first is a nerdy bit about the ease of doing state-of-the-art multi-lingual speech-to-text transcription on your own computer, and the second is about not being happy.
An applied machine learning prelude
A few months ago, I listened to a podcast with Sam Harris and Morgan Housel, titled, quite uncharacteristically for Harris podcasts, “Wealth Matters”.
Hidden between all of the money- and economics-related information, they started to talk about human well-being, and specifically about happiness vs contentment.
As I was driving, a few specific words by Sam Harris dealing with the modest but robust bliss in everyday contentment really hit home for me.
I made a mental note to find the spot in question later, and then to transcribe it for this blog.
It went the way of many of my mental notes made while driving and listening to podcasts… floating peacefully and oblivious somewhere in my brain.
… until last week, when the mental note surfaced (!!).
First I graduated the mental note into my Emacs Org mode notes. A few days later I tracked down the interview in question (man, that Waking Up app has great content, but it’s usability has much room for improvement (low hanging fruit include 1. automatically continuing current listening and 2. better search)) and made another note, which on its part led to me sitting down a bit later with my phone, trying to find the correct spot in the 2 hour podcast.
Fortunately, two attempts got me to the desired segment at 48m55s.
At that point, I could have just manually transcribed the speech in question, but it turned out to be a longer segment than expected, and then a little voice (haha) reminded me of the recent release of OpenAI’s multi-lingual text-to-speech neural network model called Whisper.
In spite of XKCD’s extremely valuable lessons on when you should or shouldn’t automate a thing, I decided that there was a learning opportunity here and so I git cloned (nerd for “downloaded”) the Whisper source code.
It took me the most of the time to get the required GPU-capable packages on
Windows, where my RTX 2070 lives (I had to revert to miniforge and conda, see
environment.yml below), and then to record the 6.5 minute audio segment from my
phone using the external microphone on my PC with Audacity (this was the only
way to get the audio out of the app).
Once that was done, Whisper took about a minute (64 seconds to be exact) to transcribe the 6.5 minutes into absolutely flawless, timestamped English text.
I was quite flabbergasted.
It is now possible to download a pretty compact neural network to your own PC, and then to use that model to transcribe automatically any number of spoken languages, optionally also translating in the process.
environment.yml for mamba
# mamba env update # conda activate whisper # whisper --language English name_of_your.mp3 name: whisper channels: - pytorch - nvidia dependencies: - python=3.10 - pip - pytorch - torchvision - torchaudio - pytorch-cuda=11.7 - pip: - git+https://github.com/openai/whisper.git
Happiness is a joke
The section that really hit home for me is the one in response to this statement by Morgan Housel, the interview guest:
That difference between happiness and satisfaction is really key and easy to overlook. Happiness, like I said earlier, is a very fleeting emotion.
I use the example of, like, if I told you the funniest joke in the world, you’ve never heard a funniest joke, you would laugh for 30 seconds, and then you’d be over it.
And if I told you that joke every day, you’d get sick of it, even if it was the best joke in the world.
Happiness is like that as well, and therefore I think the most that we can aim for, the best we can aim for, is contentment.
But when people chase happiness, I think that’s when people are setting themselves up for disappointment, because no matter how funny the joke is, so to speak, it’s a fleeting emotion that sticks around for just very brief moments before you revert to something else.
I’m not a fan of analogies, because they have the tendency to distract folks from the core of the discussion at hand. In the worst case, people think they understand the issue because they understand the analogy.
However, here we can say that the world’s funniest joke would indeed probably be as fleeting as that feeling of happiness.
I have heard people joke about Sam Harris, saying he is one of the few humans who is able to “speak in full paragraphs”.
His off-the-cuff response, the transcription of which is shown below with my emphasis, indeed seems to demonstrate that quality.
Yeah, I’m a big fan of thinking of well-being in terms of a kind of a larger footprint of contentment and equanimity and peace, you know, non-conflict, as opposed to the more transitory, hotter experiences of joy and happiness.
I mean, joy is a fleeting emotion.
It’s not to say it’s not important, and we don’t love it. But there’s a kind of a background context of just not having a problem, which is certainly underrated.
It’s not something people tend to, when they think of living their best possible life, people tend to think about joy and fun and even, you know, ecstasy, and they get bored with words like contentment or peace or equanimity.
And yet, really, when you just look at it, when you study the nature of your own psychological suffering, really, the thing that is durable, the thing that is achievable and, you know, truly enviable, if you don’t have it, is contentment, and it can deepen to the point where, you know, there’s a blissful component to it, and it’s born of not craving things that are not here in this moment, right?
I’ve mentioned at least once before my preference for contentment over happiness, although I then also cited opinions to the effect that even contentment was not possible, but I really like the way Sam so eloquently combines the achievability, the robustness, and especially the bliss that can be obtained through contentment, which itself can be cultivated through the active practice of filling the current moment with your full and deliberate presence.
P.S. about A.I.
With the current excitement over all of the applied Large Language Models (LLMS), especially around the meteoric rise of ChatGPT, I have resolved to maintain this blog as a shining example of 100% ad-free and human (namely me)-generated content.
I have nothing against applying ML, but personally I prefer my blogs as personal as possible.
It is quite poetic that after making that resolution, I did enlist the help of an ML model to transcribe the speech in the previous section.
Thinking about this some more, who knows how I might use ML in the future to help me answer questions, whilst still experiencing that it is truly me who ends up writing these words.
These distinctions are already quite blurry.
With the technology and its context evolving so rapidly, there’s no telling where this journey will take us.