TLDR:
- Launch real-time vision models by 2025 to keep pace with Google and Claude.
- Acquire new, high-quality data (possibly using robotics to gather real-world data).
- Expand hiring like Google does—offer flexible entry-level roles with a path to an elite team, aiming for a 10x growth (up to 10k employees).
- Strengthen compute infrastructure to avoid geopolitical pitfalls and remain competitive.
- Acquire small startups to secure talent (avoid a “Yahoo scenario”).
- Look to neurology research (e.g., full brain connectomes) for AGI insights.
- Reward users financially for high-quality data or evaluations.
- Monitor competition (Meta, X, Google) and consider strategic hires (e.g., Jeremy Berman).
- Strive to outpace the competition in 2025 through bigger plans and performance.
By officers I mean: Chief Execution Officer and so on. In my humble opinion: It would be interesting to have a mini DevDay with Sam, Greg and so on here, maybe one day - and if that ever happens, let’s hope most comments are about costumer service questions
I’ll start, this is what I would say:
-
realtime vision models would be nice to launch in 2025. Google already has it in their API and this is the second time since the gpt-2 release I’ve though about switching over. First time was with Claude outperforming in dev related tasks, now google is catching up.
-
Ilya is probably working on how to aquire new high quality data. The only way to not create artificially which implodes model capacity is to collect such data in the real world. My take on this is to use robotics which only purpose is capture such data in the real world. This is my main concern for OpenAI not branching into robotics
-
Look at how google is hiring, they are only going for PhDs. I’d deeply expand OpenAI hiring plans. One way I’ve succeed in a market (over 1M in profits) was to expand hires in quantity. Most of my competitiors had 30-40 employees at a high salary. I expanded the business to 300, where 30 had really high salaries and access to company more sensitive information, while the other 270 had less competitive salaries. So many would work for OpenAI at 50-80k, it would be interesting to have a program that isn’t super hard to get in but one that you can graduate from an get into the elite team. Since OpenAI is around 1000 employees now, it would be cool to see it grow to 10k employees so its able to compete with Google.
-
Sora’s team only had 6 people as far as am I aware, a 10x growth of employees there could’ve resulted in a better performance. Shout out to the dev team there.
-
Urge for a larger and more united national compute infrastructure. I know you guys are aware of this, failing this could lead to geopolitical failures. We already see this issue on voice cloning open source repositories, it would be ideal to mark this as a mission critical 2025 event.
-
Don’t be afraid to acquire small startups with talent. Remember yahoo? If you see talent and value in anyway, don’t be afraid to make offers
-
look into what neurology has going on at the edge of research, don’t be afraid to experiment and mimic. there are many labs that are pushing out high quality papers out there that almost no one is looking at. For example, the whole brain connectome for Drosophila and Ivy league labs have been pushing out data that might server as inspiration for AGI. We got what? 8 more years until the first human whole brain connectome is published?
-
reward with cash users, high quality evals or high quality chatgpt training data could become a source of income for some, financial incentives always gets the best of users
-
competition keeps an eye on this forum as well as everywhere else openai also does, make sure to invite to your team those that produce something that is valuable for all sides, for example Jeremy Berman, if you want him, so does your competition
-
careful with meta and x, they seem to be planing on collaborating in 2025. google is picking up pace. the higher you go, the bigger the competition. let’s hope this is the year openai really separates itself, in 2024 things got scary close, let’s have a better 2025. 2024 was amazing, don’t get me wrong on this, but so is your competition, big league traits require big league planing and performance.
and that all I’d have to share, maybe more if this wasn’t a public post. Now, community members, what would you say if you had the opportunity?