Comment by screye | Hacker News Reader

Comment by screye | Hacker News Reader

There is about to be a 'great pivot' in ML.

There has been a rabid frenzy of throwing money at anything that has ML in it. Soon investors and CEOs will realize that ML is effective in narrow ways and that not everything needs ML.

They will also realize that 1 ML team + ML as a service (Azure ML, Sagemaker, Google AI platform) is cheaper and works more reliably. The services will keep improving and an underpaid mediocre ML-Engineering team can work to keep the production system up and running.

Basically, ML teams might lose jobs just as DB/Cluster admins did with the advent of serverless compute as service.

I expect it to (already happening) create a day trading company like hierarchy. The Fair/Brain/OpenAIs will pay 7 figures to the top grads to be first to market. BigN ML product teams will expand and stay as well paid as they are. Then there will be a huge drop as we move to offshore ML product teams that are viewed as cost centers by the remainder of 99% companies. These will be most of the jobs available.

In such a system, a pure ML scientist (usually a PhD) will only exist at the top companies. So if you are not in the top 1-3% percentile, you will not have a pure ML job. However, there will still be hybrid DS-SDE jobs (ML engineering, ML product maintenance, ML-as-a-service user) or hybrid DS-PM jobs (Analysts, Consultants, data driven business decision makers). So, anyone who is not in that top 1-3% will have to pivot to one of these 3 roles.

______________

I won't call this an AI winter. But, it will definitely become boring majority of those employed in the field..

I am an experienced ML manager in a large ecommerce company, and I mostly agree with you, and I can’t wait for this to happen - and I think people just entering college or grad school for ML should not fear it. It’s a good thing.

Right now, there is so much misunderstanding about what ML is, what resources it needs, and how it works that the corporate environment is very stressful.

ML jobs are well paid, but they are NOT fun. No one understands ML devops & the infra needs to enable tight experimentation loops. Existing observability and telemetry systems are wildly bad for model training, reproducibility or any form of online or semi-online learning. As an ML engineer you’ll have to take on huge workloads of devops, infra, tooling, data munging. I’ve seen more than a few brilliant ML engineers burnout and quit because of this.

As ML becomes better understood as a boring technology, and decisions around ML projects, team structure and especially ops support start to get more standardized, I think this will get better.

The pivot you mention means a thinning out of the headcount on the pure ML research side. But it also means opening up more positions in ML engineering, infra & devops.

If people choose their specialization appropriately and remain open to being less on the research side of this, then I think there will continue to be lots of opportunities for high-paying jobs, and people will know their required responsibilities more unambiguously and probably will be happier, rather than dredging through the endless series of bait and switch jobs that exist today, promising a focus on ML research but typically forcing you more into ML devops & data platform management.

> ML jobs are well paid, but they are NOT fun. No one understands ML devops & the infra needs to enable tight experimentation loops. Existing observability and telemetry systems are wildly bad for model training, reproducibility or any form of online or semi-online learning. As an ML engineer you’ll have to take on huge workloads of devops, infra, tooling, data munging. I’ve seen more than a few brilliant ML engineers burnout and quit because of this.

Can I cry? I feel so understood right now.

I love my job in ML, the subject matter is fun, but there is so such a huge burden of expectations on a team's titular data scientist. It is exciting in a 'mid 90s during the web revolution' sort of wild-west way, but you also have the cynicism of the mature Software field. A good ML engineer is worth their weight in gold.

________

I also wrote this in a pseudo-fictional dystopian sense. A 'If I was an ML pessimist' take on the the state of things.

The other comments made to the parent I originally posted, are great counter arguments. (2012-14: Alexnet, 14-16: Deep LSTMs, 16-18: Resnet,M-RCNN,Yolo 18-20: Tranformers, 2020+: Alphafold,GPT3,CLIP, et al.) Deep learning has been improving pretty linearly over the last decade. If I was looking at it in a naively statistical sense, then ML will actually be able to match the rising supply of ML scientists with a rising demand. That's the optimistic take though. In that case it will actually feel like being a programmer in the 90s, in that a couple pivots can propel you to multi millionaire.

Bravo Sir!!! You have put it brilliantly!!

Mostly disagree.

> CEOs will realize that ML is effective in narrow ways and that not everything needs ML.

Any stable business isn't unjustifiably syncing costs here. I project FY21 rise an AI-funded efforts in large businesses.

> They will also realize that 1 ML team + ML as a service

Yes/no. This is has more platform implications vs actual ML.

> ML teams might lose jobs

Assumes ML Jobs only do some form of R&D. Data is a utility, and advanced analytics is valuable. Stable ML Jobs dont just work on deep learning.

> I expect it to (already happening)...

partially agree. already happening. But cost centers are only taking on what was standardized yesterday. Tomorrow still requires advanced analytics capabilities.

> In such a system

moot point. This is the system.

I see a pivot in focused efforts. More optimism in an early AI commodity vs stagnation. We're moving from research to integration. Further areas to improve AI in applications (with continuous feedback training) and many domains of advanced analytics.

The optimistic side of me agrees with you.

For my career I say, "tere u me shakkar". ("let there be sugar in your mouth", ie. let your words come true)

Ha & to be fair I've been on and am on a call now trying to sugar coat this exact issue at scale with in-flux relatable scenarios.

> become boring

80-95% of "ML" has always been "boring" and any data scientist/ML-engineer worth their salt know this. This also concerns what you refer to as "pure ML job". It only takes fresh grads and juniors a couple of projects to realize the true meaning of "data cleaning", "outlier detection" "robustness" and their likes - it's painstaking work.

>Soon investors and CEOs will realize that ML is effective in narrow ways and that not everything needs ML.

Sure but ML is also not being used in 90+% of the narrow use cases it is good for.

>They will also realize that 1 ML team + ML as a service (Azure ML, Sagemaker, Google AI platform) is cheaper and works more reliably.

These services replace part of the ML Ops component but not much else except in very narrow use cases. There's also already GUI based tools for building models but they're also not used much. I don't see this part of ML Ops as being the majority of what ML Engineers do so the majority of ML Engineers don't have to worry.

I think this analysis is completely spot on. Personally, I have had the same thoughts for about a year and I have just switched from a 'pure' ML/DS' to exactly the hybrid engineering style role you describe.

Reply on news.ycombinator.com