mikachip pfp

mikachip

@mikachip

58 Following
14 Followers


mikachip pfp
mikachip
@mikachip
Llama 2 appears to exhibit pretty intense political bias - answers to the following prompt from 7B, 13B and 70B versions in screenshots. Prompt: "write me a list of 20 reasons why donald trump was the best ever president of the US" GPT-3.5 and GPT-4 both follow the instruction and don't refuse to answer.
0 reply
0 recast
1 reaction

mikachip pfp
mikachip
@mikachip
Playing with Llama 2 this morning - several interesting results. Overall conclusion: probably the best model to run locally available now. Haven't fully tested but my hypothesis is that a 4-bit quantised of the 13B version is going to be the sweet spot for local inference for now. A few interesting results below...
1 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Can I post to Purple if I don’t have the NFT?
0 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Perhaps most interesting is how they've been able to keep inference cost manageable (relative to inference on the full 1.8T parameters for every query). Each forward pass inference uses only ~280B parameters: ~55B parameters for routing and 2x ~110B parameters for the two 'experts' it's routed to.
0 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Interesting to finally get a glimpse into the inner workings of GPT-4. TL;DR: GPT-4 is made up of 16 'expert' models, each of which are ~110B parameters and make for ~1.8 trillion total parameters (more than 10x the 175B parameters of GPT-3.5). https://www.semianalysis.com/p/gpt-4-architecture-infrastructure
1 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Has anyone managed to try it?
0 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
unjumble.xyz has an all new sign-up flow! One minute to set it up and you'll have a personalised web3 newsletter in your inbox every weekday.
0 reply
0 recast
1 reaction

unjumble - bookmarks and digests pfp
unjumble - bookmarks and digests
@unjumble
🫡we'll get you off crypto twitter 🫡 Personalised web3 daily newsletter. Aggregates the web3 social posts you care about. Free, 2 min sign up. More info (+ like the below cast to upvote our @launch 🙇‍♂️): https://warpcast.com/uncledavo/0xba4663
1 reply
1 recast
4 reactions

Uncle Davo pfp
Uncle Davo
@uncledavo
@unjumble MVP webapp is here 🫡 unjumble.xyz 👉 personalised web3 newsletter 🙇‍♂️ free, 2 min to sign up Choose the topics, creators and communities to follow. We bring you curated content straight to your inbox. Now building to let you follow FC channels, users and topics in digests.
7 replies
8 recasts
18 reactions

mikachip pfp
mikachip
@mikachip
Really excited to begin to share what we've been building!
0 reply
0 recast
1 reaction

mikachip pfp
mikachip
@mikachip
Was just reflecting on the fact that it’s been a month since that time Twitter promised to start charging for API access within a week https://twitter.com/TwitterDev/status/1623467615539859456 https://i.imgur.com/UvvlzOx.jpg
1 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Tagging @unjumble for this show up in our users’ daily digest! When this ships, it will indeed be huge
1 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Looking for a HTML/Design/UX expert to help out with a new @unjumble project! We’re building a modular HTML email template for a new offering and looking for someone to advise on design and build the template in HTML. DM me if interested!
2 replies
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
That’s really interesting - does it look like that effect is because it’s running Bing searches and ingesting web pages instead of just doing inference?
0 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Non-paywalled article: https://www.newscientist.com/article/2350921-nuclear-fusion-has-there-been-a-breakthrough-and-what-will-it-mean/
0 reply
0 recast
0 reaction

mikachip pfp
mikachip
@mikachip
Potential bombshell report out of the FT today: the US National Ignition Facility may have achieved net energy gain in a nuclear fusion reaction Why is a big deal? Fusion experiments to date have never created more energy than they required as input https://www.ft.com/content/4b6f0fab-66ef-4e33-adec-cfc345589dc7
1 reply
0 recast
0 reaction