1.AI job posting triggered the “bloodshed”
In today’s world of layoffs at major tech companies, the only thing that is not affected and is still being competed for by major companies is AI talent. Major tech giants such as Microsoft, Google, Meta, etc., are actively investing a lot of resources and time to attract and cultivate top AI talent. At the same time, many startups, research organizations, and even traditional companies, are also looking for talent with any AI background, which also makes the salary rise. But the “volume” of AI talent is not only reflected in high salaries. On the one hand, all the money is pouring into this area, on the other hand, all the best talents are also joining this field. If you are not careful, you may attract controversy. Recently, a couple of overpriced job offers for AI practitioners got into big trouble.
Netflix recently posted a job listing on its own website for a product manager for a machine learning platform, and to the shock of many, the position offered a salary range of $300,000 to $900,000 per year. That’s a salary that’s rare even in Silicon Valley, where giants go out of their way to be generous – after all, the average salary for AI-focused software engineers in the Bay Area is about $300,000, according to Levels.fyi – and the $900,000 cap Netflix is offering is just too much to pass up. At a time when the Hollywood writers’ and actors’ strike is in full swing, Netflix’s offer of this salary is undoubtedly another bucket of oil: after all, actors and writers have already felt the threat of AI, and a whopping 87% of actors are paid just $28,000 a year.
What makes Netflix offer such a high salary for this position? In the job posting, the platform develops AI that will be used to “create great content,” not just develop new algorithms to recommend shows and movies. The post also hints that the streaming giant is looking to integrate AI into “all areas of the business.” Even another product manager position on Netflix’s Machine Learning Platform team will be “gathering feedback and understanding user needs” and using AI “to optimize the production of original movies and TV shows” and ultimately help make investment decisions. This seems to imply that Netflix will start using AI to assess the funding needs of different projects. It is the union representing actors (SAG-AFTRA) that has a major concern: the growing power of algorithms to dictate the fate of people in the industry.
Representatives of Sag-Aftra claim that the algorithms will determine how many episodes are needed in a season before the number of new subscribers reaches a steady level, as well as how many seasons a series needs to run. “This will reduce the number of episodes per season to 6 to 10 and reduce the number of seasons to 3 to 4. People can’t live on that. “”The business model that’s being imposed on us is systematically squeezing our livelihoods, and it’s creating a myriad of problems for everyone up and down the line.””
With a number of people claiming that Netflix’s treatment of AI reminds them of the latest season of Black Mirror’s Joan is Awful, in which the streaming platform called Strawberry utilizes the very same AI technology to directly generate avatars of actors to automate performances based on AI-written scripts, it’s safe to say that the nightmares of Hollywood’s actors and writers are slowly turning into a reality.
This isn’t the only AI job ad Netflix is advertising with a huge salary. According to The Intercept, Netflix is also looking to hire a director of generative AI technology at its up-and-coming game studio, with a salary of up to $650,000 a year.
These efforts to “replace real people with AI” are slowly starting to bear fruit, with Netflix currently airing a Spanish reality dating series called Deep Fake Love, which scans participants’ faces to create AI-generated “deepfakes,” and its game studio, The Intercept. deepfakes” , while its game studio uses generative AI to write narrative and dialog. This all comes after the Screen Actors Guild (SAG-AFTRA) rejected an offer from the Alliance of Motion Picture and Television Producers (AMPTP) to “generously” provide actors with a one-time $200 per diem fee to have their faces scanned for future use as AI-enhanced CGI avatars forever. Augmented CGI avatars. According to AMPTP’s offer, the company would “own that scan, their image, their likeness, and be able to use it in perpetuity in any project they want, without consent or compensation.”
So that’s what Netflix’s job ads were for, and the actors and writers who were on strike blew right out of the water. But Netflix isn’t the only one playing AI math. The tech company Realeyes is also looking for actors to take on jobs: two hours of “expressing different emotions” and “improvising short scenes” in order to “train AI databases to better express human emotions,” for two hours of pay. human emotions” for $300. The post also emphasizes that this is for “research” purposes, not commercial ones, and Realeyes repeatedly stresses that there are limitations to training AI to create “expressive performances”. Of course, no one bought that argument. “It’s almost guaranteed that this ‘research’, when commercialized, will be used to create digital actors who will replace humans,” said Ben Cho, a professor of computer science at the University of Chicago. “This aspect of the ‘research’ is largely a distraction.”
2. Nvidia hires graduates byrequiring “8 capstone papers”
It’s not just Netflix that’s stirred up a hornet’s nest with a job posting, it’s also NVIDIA, but NVIDIA’s minefield isn’t too much money, it’s too many requirements. In addition to the usual requirements such as recent PhD graduation and deep understanding of large language models and deep learning, the job posting also puts a very explosive requirement that directly makes major researchers cursing – “At least 8 top conference (CVPR, IEEE, NeurIPS) papers. This is the requirement for NVIDIA attracted “group ridicule”. Many people couldn’t believe that this was the “minimum requirement” for new graduates in the machine learning field, and others said, “Why don’t you list the Turing Award as well?
At first, Mark Moyou, a senior data scientist at NVIDIA, tried to explain. He said, “If you’re doing fundamental research in LLMS, you have to have a very specific set of skills, so I’m not surprised that the requirements are so specific. The optimization work that these guys have to do is very high level, it’s low kernel optimization and it’s very fast paced. It also has to be done on the largest scale computations.” Most people didn’t buy his explanation, though. Some said that eight papers is not a very specific skill set, and that such a number was likely designed for a particular candidate who had already been internalized. Others designed a short paragraph:
HR: Isn’t it a bit much to ask a PhD candidate to publish papers?
TeamLead: But my buddy has 8 papers!
HR: Why don’t we round it up to 10?
TeamLead: That’s too many. 8 is just right, that’s it, and adding buzzwords to it, it has to look like we’re doing something cool!
Someone even dug up Mark Moyou’s Google Scholar page and said he doesn’t even qualify. Not to mention this guy, this whole group or even the head himself doesn’t measure up. Others say that even OpenAI co-founder Andrej Karpathy doesn’t measure up, not to mention NVIDIA itself – after all, he’s only got seven papers to his name at the top of the conference.
However, some people say, NVIDIA this requirement does not say that you need to be 8 a work, if you work with others more, this is still possible. But this is still a very strange request, if someone else has 7 high-quality, and another has 8 low-quality? It was also mocked that maybe the job posting was generated by NVIDIA’s own big language model, which didn’t seem to be of good quality. Seeing the ridicule, NVIDIA took the posting down, apologized that it was an error, and reposted the requirements for applying for the position, replacing at least 8 top papers with “a strong track record of presenting papers at top papers”. But judging from Twitter, people aren’t buying it. It seems that this “roll up to the sky” of the job announcement, the AI researchers who are already feeling the pressure of competition to offend deeply.
3.Sky-high salary, creative cake
There is so much money on the market, but the talent is so little, have no choice, more and more companies began to use a variety of creative ways to attract talent to join. This includes OpenAI’s unique salary program. According to Levels.fyi, OpenAI offers a median salary of about $925,000 to software engineers. That includes a base salary of $300,000 and $625,000 in PPUs, a new type of financial incentive called Profit Participation Units (PPUs).
The core value of PPUs is the ability to share in the profits generated by OpenAI. This means that those who hold PPUs can take dividends on them when OpenAI makes money, or they can sell them to other investors. It sounds like PPUs are similar to stocks.
However, Sam Altman has repeatedly emphasized that OpenAI will not go public in the future. This also means that the value of PPUs is determined by the company’s internal guidelines and valuation tools. In short, it’s up to OpenAI itself. And relative to stocks, the ultimate value of PPUs seems much more uncertain: PPUs are based on OpenAI ultimately being able to achieve profitability. If it fails to turn a profit, there is a risk that the value of the PPU will go to zero. For an organization whose stated goal is to spend billions of dollars developing general-purpose AI, this could take years.
If they are sold to a willing buyer, PPUs are currently capped at 10 times their original stated value due to OpenAI’s structure as a “profit cap” company. This means that a PPU worth $1 million can only be sold for a maximum of $10 million. Additionally, OpenAI’s PPUs have a two-year lock-in period, which means that new hires must keep their equipment for two years before they can sell it. But for the engineers who are now crowding in to join, who really cares if the PPUs are capped at 10x or 20x under the aura of OpenAI?