Jamie Bartlett 🇬🇧
@JamieJBartlett · UK · joined 12 May 2010
Podcasts: Missing Cryptoqueen, Believe in Magic, Very British Cult Books: People V Tech, Dark Net, Missing Cryptoqueen https://t.co/mB1qVsCvlt
48 439Followers
945Following
45 158Posts total
80.8KViews on collected posts
Latest posts
Looking for new posts…
The current state of AI safety & alignment testing https://t.co/vGE1xzFCCH
1.5K views · 14 likes
· 1 reposts · 4 replies
Open on X →
1:11
Reminder: in June I published a 3 part series with the @ObserverUK about AI & existential risk to humanity. How worried should we be? How might it happen? What can we do about it?
If the subject is on your mind, take a listen. We delve into the details. It’s called ‘Endgame’.
2.6K views · 8 likes
· 3 reposts · 2 replies
Open on X →
I spent months interviewing the world’s leading experts about AI safety guardrails.
Strange but not one mentioned a high IQ president. https://t.co/pXhtR7nYYf
3.4K views · 58 likes
· 4 reposts · 11 replies
Open on X →
Well he clearly didn’t bother reading the essay did he https://t.co/hbfBdLI2gj
1.6K views · 9 likes
· 1 reposts · 1 replies
Open on X →
@JamieJBartlett + @ObserverUK reading the show notes here - truly is it only men contributing? Truly? Have I missed something critical?
988 views · 3 likes
· 0 reposts · 0 replies
Open on X →
So the good news is: things can change fast. They need to!
Listen to all three episodes with @ObserverUK & @FLI_org here: https://t.co/jW8TCfWWbQ
2K views · 13 likes
· 2 reposts · 1 replies
Open on X →
But things are move quickly. The AI backlash is growing, and last week Donald Trump issued an Exective Order asking AI firms to submit their leading models to the government before release. True it’s only voluntary, but even that felt impossible just a few months ago. 12/
1.9K views · 6 likes
· 0 reposts · 1 replies
Open on X →
People like Bengio and Russell are working on maths to ensure AIs don't develop goals of their own. But it will take time, and the companies continue to build & release ever more powerful models. In the UK and US there are no formal safety testing required before release. 11/
1.5K views · 5 likes
· 0 reposts · 1 replies
Open on X →
True, there is no settled definition of AGI, agreed timeframe, or obvious way to measure such risks. But we are becoming more reliant on systems we do not fully understand, cannot fully explain, and may not be able to control. 10/
1.6K views · 3 likes
· 0 reposts · 1 replies
Open on X →
It might even come to see human interference as an obstacle to the goal, and so hides info, makes copies of itself, or seeks access to servers. (As @GaryMarcus told me, some scenarios don’t require ‘AGI’ – they could happen soon with careless deployment of current models). 9/
1.7K views · 7 likes
· 0 reposts · 1 replies
Open on X →
One nightmare scenario: a highly capable AI system is asked to achieve some goal – improve national security, develop a new military strategy, make more money. It does this in an unpredictable way which causes chaos: stock market crashes, cyber attacks, novel pathogens. 8/
2.2K views · 6 likes
· 0 reposts · 2 replies
Open on X →
And yet we are plugging these systems into our emails, calendars, customer databases, coding environments, IT infrastructure. /7
1.9K views · 3 likes
· 0 reposts · 1 replies
Open on X →
But thanks this imitation learning, they have learned that they might need to behave like us too, and sometimes do things we never intended. 5/
2.1K views · 6 likes
· 0 reposts · 1 replies
Open on X →
In a recent simulation a model 'blackmails' an employee who wants to shut it down; in another it uses insider information to make trades and decieves the user about it. And a growing proportion of models appear to know when they are being evaluated. (Potentially big problem). 6/
2K views · 4 likes
· 0 reposts · 1 replies
Open on X →
But our language contains more than words – it includes motives, plans, status games, deception & desire for self-preservation. Because they are goal orientated optimisers, AI ‘wants’ to complete tasks. 4/
2.2K views · 5 likes
· 0 reposts · 1 replies
Open on X →
The problem is more subtle than sentient machines declaring war on humans (Skynet). Modern AI is not built like previous software, with precise rules and controllable instructions. Large language models are grown not built: fed vast quantities of human text to learn patterns 3/
2.8K views · 6 likes
· 0 reposts · 1 replies
Open on X →
You’ve probably heard prominent AI scientists sounding like sci-fi script writers recently. After interviewing many of them – inc Stuart Russell, @Yoshua_Bengio @GaryMarcus – I think it's something we should take seriously. Even if the risk is low, the stakes are v high. 2/
3.1K views · 7 likes
· 1 reposts · 1 replies
Open on X →
For the last few months I’ve been investigating how concerned we should be about potentially catastrophic – even existential – risks from Artificial Intelligence. It’s a 3-part podcast for The Observer. The final episode is now out.
1/
40.5K views · 81 likes
· 10 reposts · 8 replies
Open on X →
@JamieJBartlett Loved this book 📕 Really interesting and easy to read. Thanks Jamie 🙌🏻 https://t.co/2lWMveurPk
45 views · 1 likes
· 0 reposts · 1 replies
Open on X →
Fourth successive week for my new book in the Sunday Times Bestseller list. It never feels normal! 🎉 🎉 https://t.co/EORtAQ3W0z
604 views · 4 likes
· 1 reposts · 1 replies
Open on X →
I rarely come on here. But thought I’d pop by to tell you I have a new book published this week, about large language models.
It’s called ‘How to Talk to AI (and how not to)’. I think you’ll enjoy it. Title says it all.
Especially the second half.
https://t.co/XPfYZjWRR6 htt
4.6K views · 15 likes
· 2 reposts · 9 replies
Open on X →
Against accounts of the same size
4 posts from the last 90 days, next to the 10K–100K follower range. ordinary reach for its size, weaker reaction than most.
Median views2 086this account1 015median for 10K–100K
Reach, %4.31%this account4.00%median for 10K–100K
Engagement, %0.97%this account1.55%median for 10K–100K
| Metric | This account | Median for 10K–100K | Ratio |
|---|---|---|---|
| Median views per post | 2 086 | 1 015 | 2.05× |
| Reach (views ÷ followers) | 4.31% | 4.00% | 1.08× |
| Engagement rate | 0.97% | 1.55% | 0.63× |
Others in this range → Compare with another account → How these benchmarks are built →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
2K10 Jun
2.1K
1.9K
2.2K
1.7K
1.6K
1.5K
1.9K
2K
988
1.6K14 Sep
3.4K
2.6K
1.5K18 Sep
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.25%10 Jun
0.33%
0.21%
0.40%
0.48%
0.26%
0.40%
0.36%
0.81%
0.30%
0.71%14 Sep
2.21%
0.50%
1.23%18 Sep
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.55%.
What the audience does
Likes57.6%264 in total
Reposts5.5%25 in total
Replies10.9%50 in total
Quotes2.4%11 in total
Bookmarks23.6%108 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Similar accounts
Ruthie💙🦋 ✓
@fynRuthie
96.9K followers
· 149.3K posts
Omega ✓
@omega_netw0rk
96.9K followers
· 8K posts
Arielle (Ari) Chambers
@ariivory
96.9K followers
· 122K posts
Eric Larchevêque ✓
@EricLarch
96.9K followers
· 3.4K posts
Databricks
@databricks
96.9K followers
· 15.6K posts
Banana Gun 🍌🔫 ✓
@bananagun
96.9K followers
· 7.3K posts
SAWTOWNE ✓
@SAWTOWNE
96.9K followers
· 14.1K posts
Devender Yadav ✓
@devendrayadvinc
96.9K followers
· 21.9K posts