Fascinating analysis of AI’s effects on freelancer hiring demand: everyone wants a chatbot

Bloomberry Jobs Report

tldr: writing, translation, & customer service jobs declined; everyone wants a chatbot, apparently

A few excerpts:

To my pleasant surprise, most of the job categories actually had an increase in the number of jobs since ChatGPT was released, with the exception of 3 categories that had large declines in jobs.

The above is important to keep in mind for AI luddite doomsayers; as the history of technology has repeatedly shown, though there may be short term disruption and harm to labor, the long term trend is that technology = more (newer and better) jobs

I was surprised to see graphics design, video editing/production, and even software development jobs go up, given all the anecdotal stories we’ve been hearing about people using ChatGPT to generate code, illustrations and even full featured videos.

And surprising on first blush, although perhaps logical if you assume good general purpose AI has reduced the need for specialized ML:

While there was a initial increase in data annotation jobs, the # of data annotation jobs have been fairly flat in the past 10 months. And the # of machine learning jobs has actually decreased a bit since ChatGPT was released.

Finally, I guess everyone loves to chat:

The most popular use case, by far for AI right now is in developing chatbots

Full article: https://bloomberry.com/i-analyzed-5m-freelancing-jobs-to-see-what-jobs-are-being-replaced-by-ai/

Random facts – things I learned (March 8 2024) – “if you make an effort in training when you don’t especially feel like making it, the payoff is that you will win games when you are not feeling your best”

RANDOM FACTS

I think he’s a genius because he’s solved a huge problem, which is that writing is really lonely. And it’s brutal to sit down and be by yourself and write. He eliminated it, first, by the dictation, by having the typist. And now he’s totally eliminated it by having a roomful of people. It’s a hell of a lot more fun than sitting alone in a room and feeling depressed

Like, wow, an AI that can write a Reddit comment! Well, there are millions of Reddit comments, which is precisely why we now have AIs good at writing them. Wow, an AI that can generate music! Well, there are millions of songs, which is precisely why we now have AIs good at creating them.
Call it the supply paradox of AI: the easier it is to train an AI to do something, the less economically valuable that thing is. After all, the huge supply of the thing is how the AI got so good in the first place.

Policymakers want wealth-flation (assets go up) but not pleb-flation (because that leads to unrest and protest)

More recent efforts to optimize The Pile (and its relatives) for language model training arrived at the same conclusion: more web text makes the model smarter. This is counterintuitive: doesn’t the median quality level of web text pale in comparison to hand-picked high quality text corpora? The answer seems to be diversity: web text, for all its failings, has nearly every conceivable usage of language

In May of last year, Andrej Karpathy tweeted his view of the sufficient conditions for good datasets, and so strong models: ‘Large, Clean, Diverse’.

jobs like generating AI content, developing AI agents, integrating OpenAI/ChatGPT APIs, and developing AI apps are becoming the rage. But by far the #1 use case? Chatbots, with the # of jobs related to developing chatbots exploding 2000% since the release of ChatGPT and the OpenAI API. If there is a killer use case for AI today, it’s in developing chatbots.

There is a long tradition of this: The first automobile (pictured above) looked like a horse-drawn carriage without the horse, early telephones looked like telegraph systems, early movies looked like filmed plays.
YouTube became one of the biggest winners of Web2 because it broke this skeuomorphic mold by being the first video-hosting service to go all-in on user-generated content.

To my pleasant surprise, most of the job categories actually had an increase in the number of jobs since ChatGPT was released, with the exception of 3 categories that had large declines in jobs.
The 3 categories with the largest declines were writing, translation and customer service jobs. The # of writing jobs declined 33%, translation jobs declined 19%, and customer service jobs declined 16%

A single layer perceptron (SLP) is a feed-forward network based on a threshold transfer function. SLP is the simplest type of artificial neural networks and can only classify linearly separable cases with a binary target

But, then, perhaps the Genoan was like those clever men who never know more than they need and believe only what it is in their interests to believe

The thrill of winning is in direct proportion to the effort I put in before. I also know, from long experience, that if you make an effort in training when you don’t especially feel like making it, the payoff is that you will win games when you are not feeling your best – Rafa Nadal’s memoir

“With GrubWithUs we learned that friction can kill marketplaces. On eBay, you could have 300 sellers who were listing the exact size of shoe you were looking for, and you would have to sort through ratings, comments, shipping information to find exactly what you need. We wanted to remove that friction from the experience. And since we’re an authenticated marketplace, people could trust that they would receive exactly what they ordered,” says Lu.

Pluto takes 247.94 Earth years to orbit the Sun. According to my calculations, the Plutonian year that started on July 4, 1776 will end this year on June 12, 2024

Google groups users based on their past behavior to predict what they want. Think about it like Amazon’s “other shoppers also bought”. Multiplied by hundreds of billions of searches, strong patterns emerge

The heavier an animal, the easier it dies from a fall. You can drop an ant over 15,000x its height (~1,250 feet) and it won’t die. Squirrels can be dropped 150x their height. Humans die around 10x our height. If you drop an elephant just 1x its height (~10 feet) it dies.
The bigger you are, the harder it is to reproduce, gestation times take longer. Ant eggs hatch within a couple weeks of being laid. Humans take 9 months. Rhinos 17 months. Elephants take nearly 2 years.

When you read biographies of people who’ve done great work, it’s remarkable how much luck is involved. They discover what to work on as a result of a chance meeting, or by reading a book they happen to pick up. So you need to make yourself a big target for luck, and the way to do that is to be curious. Try lots of things, meet lots of people, read lots of books, ask lots of questions.

just get 1% better at whatever you’re working toward each day and you’re guaranteed to make progress

…there is a pretty good correlation between those who work with the doors open and those who ultimately do important things, although people who work with doors closed often work harder. Somehow they seem to work on slightly the wrong thing – not much, but enough that they miss fame – PG

There are five types of time:
1. Micro Time (sub-second)
2. Engagement Time (Seconds)
3. Business Time (Minutes to Hours)
4. Strategy Time (Days to Weeks)
5. Big-Thinking Time (Months to Years)

Bill Wilson, the founder of Alcoholics Anonymous: “You can’t think your way to right action; you can only act your way to right thinking.”

“I wish it need not have happened in my time,” said Frodo.
“So do I,” said Gandalf, “and so do all who live to see such times. But that is not for them to decide. All we have to decide is what to do with the time that is given us.”

In practicing a skill in the initial stages, something happens neurologically to the brain that is important for you to understand. When you start something new, a large number of neurons in the frontal cortex (the higher, more conscious command area of the brain) are recruited and become active, helping you in the learning process.

The least and most successful among the Italian Americans were the most ardent admirers of Mussolini’s revolution; the least and most successful among the Irish Americans were the most responsive to De Valera’s call; the least and most successful among the Jews are the most responsive to Zionism; the least and most successful among the Blacks are the most race conscious.

Your body is in elimination mode in the morning and drinking water kick-starts your body’s functions and assists in the elimination process. This helps you to feel energised and replenished ready for a great day

Tom Hanks on acting: Hit the marks and tell the truth!

out of the three core layers of internet stack – naming (DNS), transportation (TCP/IP) and application (HTTP), naming is at the very start of the stack

On February 22, 2024, the closing price of the Nikkei Stock Average surpassed the record high of 38,915.87 yen set on December 29, 1989, the peak of the bubble economy. This was the first time in 34 years

Scientific literature shows that adults who exercised for at least 30 minutes a day slept an average of 15 minutes longer than those who did not exercise [19]. Other studies have shown that physical activity can help to reduce sleep disorders, such as insomnia, daytime sleepiness, and sleep apnea [15,19,20]

Amateurs have a goal. Professionals have a system.

Doing the same thing every day seems easy but it actually isn’t

If you don’t believe in God, it may help to remember this great line of Geneen Roth’s: that awareness is learning to keep yourself company. And then learn to be more compassionate company, as if you were somebody you are fond of and wish to encourage.

What does “tape out” mean in chip manufacturing?
The term “tape-out” is used in chip design to signify the completion of the design phase and the start of manufacturing. Originating from older days when the final design was written onto magnetic tape and sent out for fabrication, the term has endured as a symbolic milestone

I should make sure that I’m sufficiently exhausted from working that no one can keep me up at night. That’s really the only thing I can control. – Jensen Huang

We’ve since come to understand that actual biological neurons are substantially more complex than our early models of them, and neural networks have virtually no similarities to the design of actual brains. For instance, the common locust uses a single neuron for implementing its collision detection while flying. This is done through complex dendritic integration, which cannot be represented by our simplified neuron models with linear summation.

Participants who were genetically biased not to have a tend-and-befriend response got the biggest health benefit of being prosocial. The scientists speculated that caring for others can jump-start the oxytocin system, even if you have a genetic predisposition that makes a tend-and-befriend response less likely.

A great surprise that emerges from the genome revolution is that in the relatively recent past, human populations were just as different from each other as they are today, but that the fault lines across populations were almost unrecognizably different from today.

I think as technical people we have a strong bias to put up code or papers or the final thing and feel like things are mostly self-explanatory. It’s there, and also it’s commented, there is a Readme, so all is well, and if people don’t engage then it’s just because the thing is not good enough. But the reality is that there is still a large barrier to engage with your thing (even for other experts who might not feel like spending time/effort!), and you might be leaving somewhere 10-100X of the potential of that exact same piece of work on the table just because you haven’t made it sufficiently accessible

Morgan Housel gets his best ideas while walking. In an interview, he said, “If I ever get some sort of writer’s block, or I’m just trying to think an article through, I go for walks. I go for two or three walks per day, and that’s where all of the writing happens, and I usually take notes when I walk.”

I think the most interesting part of the paper is the finding that walking improves creativity not due to environmental stimulation, but due to walking itself. Whether outdoors or on a treadmill, walking improved the generation of novel and appropriate ideas. Surprisingly, this effect extends to sitting after a walk

One ought to go too far, in order to know how far one can go.

I don’t, for the record, think we are at an iPhone moment when it comes to virtual reality, by which I mean the moment where multiple technological innovations intersect in a perfect product. What is exciting, though, is that a lot of the pieces — unlike three years ago — are in sight. Sora might not be good enough, but it will get better; Groq might not be cheap enough or fast enough, but it, and whatever other competitors arise, will progress on both vectors. And Meta and Apple themselves have not, in my estimation, gotten the hardware quite right. You can, however, see a path from here to there on all fronts

Prior editions:

The killer app comes first (in both crypto and AI)

In crypto there’s a constant chicken and egg debate of apps versus infrastructure. Which is more important? Which accrues more value? As an entrepreneur, which should I build?

For me, this 2018 article effectively settles the debate: https://www.usv.com/writing/2018/10/the-myth-of-the-infrastructure-phase/

The answer: It depends what part of the cycle we’re in.

But when you look at the history of general technologies, the killer app comes first. The infrastructure follows.

For example, light bulbs (the app) were invented before there was an electric grid (the infrastructure). You don’t need the electric grid to have light bulbs. But to have the broad consumer adoption of light bulbs, you do need the electric grid, so the breakout app that is the light bulb came first in 1879, and then was followed by the electric grid starting 1882. (The USV team book club is now reading The Last Days Of Night about the invention of the light bulb).

You could say a series of technological breakthroughs (eg, the right filaments, the right glass container) enabled the first “killer app” (💡💡💡) which then incentivized the infra.

Another example:

Planes (the app) were invented before there were airports (the infrastructure). You don’t need airports to have planes. But to have the broad consumer adoption of planes, you do need airports, so the breakout app that is an airplane came first in 1903, and inspired a phase where people built airlines in 1919, airports in 1928 and air traffic control in 1930 only after there were planes

Same pattern here: a series of new technologies (lightweight engines, proper control mechanisms) enabled the first “killer app” (🛫🛫🛫) which then incentivized the infra.

Crypto’s first killer app is right under our noses: Bitcoin itself.

The killer app was Bitcoin! And what it represents: a sovereign store of value tied to an uncensorable payment network.

Satoshi’s technology breakthrough enabled the killer app (Bitcoin) which has now enabled more than a decade of crypto infrastructure buildout, from alternative Layer 1s to smart contracts to new blockchain primitives.

In generative AI, I think a similar pattern is also unfolding:

ChatGPT was the first AI killer app. The lightbulb moment. 100M+ users within months of launch and one of the fastest growing consumer apps of all time.

ChatGPT opened investors eyes’, blew users’ minds, and now everyone from Google to Softbank to the CCP are spending billions ($7 trillion??) to build and buy AI infrastructure.

And steadily and surely, much of this infrastructure investment and innovation will make AI better, faster, and cheaper. Then more killer apps will be built atop all the GPUs, foundation models, and SDKs. Which then begets more infra. And the cycle continues.

Collection of recent crypto learnings 3: “…one can argue that currencies themselves are intrinsically platforms, and that coexisting multiple currencies should be analyzed as platform competition.”

Past updates 1 and 2

The ENS approach is even more vulnerable, where a group of multisig key holders, no matter how reputable, will control the governance and upgrade of the backbone infrastructure of the decentralized web.

I believe, in reality, a significant portion of the cryptocurrency space operates on meme culture,” Zhu said during the AMA. “We all tend to invest in bitcoin because it represents something everyone believes in, transforming it from a meme into a tangible reality

out of the three core layers of internet stack – naming (DNS), transportation (TCP/IP) and application (HTTP), naming is at the very start of the stack

Good breakdown / categorization of AI+crypto projects


(But missing generative media like images, videos)

the core definition of a blockchain is all the data used is generated within that blockchain and therefore verifiable by any participant in the blockchain. Towards that same end, smart contracts can only talk to smart contracts

Furthering the idea that the US has much to gain from the adoption and co-option of Bitcoin is the tangible stash of coins distributed within its borders; MicroStrategy’s 189,150 bitcoin, the 215,000 bitcoin seized by the Department of Justice, Block.one’s 164,000, Grayscale’s 487,000 in GBTC, and now the new US spot ETF offerings hold a combined 170,174 bitcoin as of 1/31

in a 2011 interview with Bloomberg, Fink went so far as to say “Markets don’t like uncertainty. Markets like, actually, totalitarian governments… Democracies are very messy.”

Bitcoin is punk rock. You don’t get it? Fuck you we don’t care. We’re having a party — Peter McCormack

CDixon (paraphrased): “Computer” is a collection of both nouns and verbs. A ledger is just a noun. So it undersells the power of verbs like earn, transfer, spend, save, stake, lend, etc

From an app’s perspective, blockchains offer three key features: consensus, composability, and availability 🧵
1. consensus – solve contentious race conditions
2. composability – access other liquidity and apps
3. availability – data is readily accessible
// what about governance (consensus?), tokenomics (new biz model)

The Ethereum blockchain core developers did briefly consider including an ALARM opcode to enable smart contracts to schedule operations in future blocks, but it was ultimately discarded as unworkable [https://vitalik.ca/general/2022/03/29/road.html]. The Cosmos SDK used for development of application specific blockchains [https://v1.cosmos.network/sdk] has some support to execute code – with significant limitations – at the beginning and end of each block.

Blockchains invert the hardware-software power relationship, like the internet before them. With blockchains, the software governs a network of hardware devices. The software—in all its expressive glory—is in charge.

one can argue that currencies themselves are intrinsically platforms, and that coexisting multiple currencies should be analyzed as platform competition.

That said, when a token goes straight down, you can’t call this a screaming success. There is a good reason why IPOs generally go up. And there is a good reason for why BNB, ETH, and BTC are 3 of the most successful protocols today. When you price an asset low, and let early investors participate in the financial upside of your success, it tends to have long-lasting positive effects. Your users become power users and evangelists. But when something prices high and goes straight down, you alienate those who were true believers. And it’s hard to come back from that

AI+blockchains point to a dystopia of impersonal and faceless interchangeable-parts humanity that’s more industrial than the industrial age.

Why not put $500 into a memecoin that could 50x, knowing that you could likely lose most or all of it? It’s not like the $500 is enough to make any difference anyways. Neither is $1k or $5k. That mindset, which is becoming pervasive in America, is financial nihilism. This is the zeitgeist for young Americans, you’re naïve to think otherwise. And it’s a huge driver of shitcoining

For Web 3 to succeed it needs to do two things:
Enable cool functionality unable through traditional Web 2,
and make the user largely unaware that they’re even on the blockchain

Programmable, composable data structures (ie, tokens) are the “new computing primitive” that will usher in the next phase of the internet

We need an alternative. Crypto is the perfect marriage for AI since the transparent global human coordination that underpins the movement is something that can harness AI for good at global scale. Crowdfunding (with cash or with your GPU) the creation and fine tuning of open source models which anyone can audit in real time for biases or issues is the safest path forward in the accelerating world of AI.

The idea of Bitcoin, like the idea of Index funds is a clean “world view” that markets itself. It’s not the only crypto that does so. Once you do accept Bitcoin into your brain, part of your brain opens up to other cryptos: Eth, Solana, NFTs, Ordinals … maybe some combination of Crypto and AI like Tao.

My sense is that this new idea: Bitcoin, and this new demographic: Millenials are in for an epic bull run.
The BTC ETF will be the gateway drug for this. It will get the Boomers and GenXs so that they CAN participate in the transition. Most won’t. But enough will.
It’s an idea that will take over the next 20 years.

USDT on the Ethereum network shows an average transfer size of $35,000, indicating its involvement in substantial financial activities within the DeFi ecosystem, likely influenced by Ethereum’s higher transaction fees. Conversely, USDT on the Tron network presents a distinct scenario. With Tron’s minimal transaction fees, the average transfer size for USDT is around $7,000, facilitating more frequent, lower-value transactions

He defines crypto as a meeting of “generative tech” (the creation of new things, users and markets) and “participatory capital formation” (individuals pooling money in new ways to create new types of businesses).

Truly valuable technology trends toward free and ubiquitous (another Kevin Kelly read)

This one’s also going in the personal bible archives

Original source: https://kk.org/thetechnium/technology-want/

Some excerpts:

“there has been a downward trend in real commodity prices of about 1 percent per year over the last 140 years.” For a century and half prices have been headed toward zero.

GPS was a novelty luxury only a few years ago. It was expensive. As its technical standards spread into mapping services and hand helds, it becomes essential, and the basic service (where am I?) will become a commodity and free. But as it drops toward the free, hundreds of additional advance GPS functions will be added to the fixed function so that more people will pay ever more for location services than anyone pays now. Where-am-I information will be free and ubiquitous, but new services will be expensive at first.

As crackpot as it sounds, in the distant future nearly everything we make will (at least for a short while) be given away free—refrigerators, skis, laser projectors, clothes, you name it. This will only make sense when these items are pumped full of chips and network nodes, and thus capable of delivering network value.

Automobiles, like air travel, are headed in direction where all software and digital devices are headed: toward the free. Imagine, I said, if you could give away a very basic no-frills car for free

A car will move you from A to B, but it also offers privacy, immediacy of travel, a portable office, an entertainment center, status, and design joy

Google has the same opportunities with them that all producers have. They offer free commodities and charge for premium services. Search is free; yet they charge enterprises for custom Google search. Or they shift their customer from reader to advertiser; in Google’s eyes the chief audience for search is advertising companies, whom they charge

Technology wants to be free, as in free beer, because as it become free it also increases freedom. The inherent talents, capabilities and benefits of a technology cannot be released until it is almost free. The drive toward the free unleashes the constraints on each species in the technium, allowing it to interact with as many other species of technology as is possible, engendering new hybrids and deeper ecologies of tools, and permitting human users more choices and freedoms of use