Joshua Achiam
@jachiam0
Followers
23K
Following
15K
Media
324
Statuses
7K
Freedom, flourishing, and abundance. Head of Mission Alignment at @openai. Main author of https://t.co/cKuSh21yaz
Joined February 2018
Excited to share what I've been working on these last few months! Spinning Up in Deep RL, an educational resource that we hope can bring down the barrier to entry in RL. https://t.co/ziYOSK6uyZ
17
252
880
Aardvark is changing how we discover and fix vulnerabilities at scale. It's truly a glimpse into the future of cybersecurity. Congrats on the launch @embeddedsec!!
Now in private beta: Aardvark, an agent that finds and fixes security bugs using GPT-5. https://t.co/xwtJhfDM3X
2
10
82
Now in private beta: Aardvark, an agent that finds and fixes security bugs using GPT-5. https://t.co/xwtJhfDM3X
210
372
3K
there are many studies flying around and sure to be many more, but this one most reflects what we see on the ground at openai ai is beyond the experimentation phase in enterprise - 95% of respondents here have used or intend to use chatgpt for work over time, process will
👀New data on the corporate ROI from generative AI from a large-scale tracking survey by my colleagues at Wharton. They found that 75% already have a positive return on investment from AI, less than 5% negative return. Also 46% of businesses leaders now use AI daily themselves.
12
21
228
Now in research preview: gpt-oss-safeguard Two open-weight reasoning models built for safety classification. https://t.co/4rZLGhBO1w
openai.com
New open safety reasoning models (120b and 20b) that support custom safety policies.
160
290
2K
Always curious and full of energy, Beagles turn every walk into an adventure 🐾🎉.
147
620
10K
Very interested in the strongest takes in either direction on this
0
0
3
We really only get to pick one or the other. A middle ground is an inherently temporary state. Each option has some strategic advantages and some strategic drawbacks, neither seems purely and obviously correct to me right now.
2
0
5
On chip exports: this is an observation, not a position. It seems like either 1) China gets our advanced chips, which creates interdependence and we get signals about their compute, or 2) they develop chip independence and there is opacity about their compute capabilities.
8
1
37
We completed our recapitalization. The non-profit, the OpenAI Foundation, is now one of the best resourced philanthropies ever, with equity valued at ~$130B. It continues to control the OpenAI for-profit, which is now a public benefit corporation. https://t.co/TevJDA3QwB
openai.com
By Bret Taylor, Chair of the OpenAI Board of Directors
743
433
4K
But it is the people who build the practicable policies who will have the greatest impact on achieving AI safety policy outcomes, and the field should figure out as much as possible how to embrace those relatively rare individuals.
2
1
26
That feeling when ya already lost everything and you watch em all crash
4
4
10
and people who are idealistic and principled but who rally behind extremely impractical banners. There is an exhaustion that sets in for anyone who cares about AI safety but doesn't align in the details with the ultra-high-risk-model crowd.
3
0
25
.@deanwball is making important, epistemically sound contributions to the AI safety/policy discourse and there is a genuine imbalance between people like him who want policy that is practicable and achieves the intent,
4
3
73
We’re very focused on making GPT-5 safer, and continue to make a lot of progress:
openai.com
We worked with more than 170 mental health experts to help ChatGPT more reliably recognize signs of distress, respond with care, and guide people toward real-world support–reducing responses that...
150
23
315
We’ve updated the OpenAI Model Spec – our living guide for how models should behave – with new guidance on well-being, supporting real-world connection, and how models interpret complex instructions. 🧵
77
15
129
🧵Today we’re sharing more details about improvements of the default GPT-5 model in responding to sensitive conversations around potential mental health emergencies and emotional reliance. These changes reflect the careful work of many teams within OpenAI and close consultation
Earlier this month, we updated GPT-5 with the help of 170+ mental health experts to improve how ChatGPT responds in sensitive moments—reducing the cases where it falls short by 65-80%. https://t.co/hfPdme3Q0w
242
71
587
Today we shared an update about recent improvements to how ChatGPT responds in sensitive moments. We also published an update to the Model Spec to make some of our longstanding safety goals explicit. This is ongoing work and we’ll continue to share updates.
🧵Today we’re sharing more details about improvements of the default GPT-5 model in responding to sensitive conversations around potential mental health emergencies and emotional reliance. These changes reflect the careful work of many teams within OpenAI and close consultation
141
14
134
I never know what's going to get a big response but I was sure this one was not it. 234 bookmarks. Wow.
0
0
10
New report exposes the truth behind the US-brokered “peace” deal in the DRC: it’s not about peace – it’s about controlling Congo’s minerals. Rwanda becomes a processing & exporting hub while Congolese communities continue to pay the price of extraction. Read how foreign
0
13
20
A critical ingredient in rebuilding American manufacturing is to tell manufacturing stories. Give me just one nationally-famous prestige drama about industrial competition where the development of factories, equipment, and distribution is the main plot, I'm begging you
we need to compete with the Chinese on the category of bureaucratic realism in fiction
3
1
47
I am a little disappointed at how rarely dot-product attention mechanisms are explicitly described as differentiable lookup tables. It seems like it would just be an easier way to convey the intuition
35
23
749
It feels slightly like being in a dream to read a front-page article on Politico and wonder if I detect the faint aroma of GPUs humming. The world is changing in subtle ways.
2
0
54
What the outside world largely does not understand about OpenAI is how decentralized and bottom-up it is, and why this is incredibly good and important for the democratization of benefits. Aidan explains it well.
@nearcyan i think your error here is thinking sora and adult content are some leadership master plan; that sama sat down with accountants and signed and said “it’s time to break glass for emergence revenue” no. i know the exact people who pushed for sora, they’re artists who worked
15
10
201
Spotlight confessions: In the beam, we shine. In the black, we dream. But what if the unseen holds the real script?
1
0
4