#2024 Mozilla Builders Accelerator: Local AI

1 messages Ā· Page 1 of 1 (latest)

fickle crow
#

2024 Mozilla Builders Accelerator: Local AI

dusk viper
#

This looks amazing!

rigid moth
#

awesome initiative ... going to share it in my circle, I know afew groups that are workign in this space.

graceful orchid
#

This is so cool. Def gonna apply for somethin

lyric solstice
#

This would be amazing. Who do we speak to?

graceful orchid
graceful orchid
lyric solstice
#

I didn’t see it in the FAQ but is this limited to one project?

merry ravine
# lyric solstice I didn’t see it in the FAQ but is this limited to one project?

The application asks for other ideas you considered applying with, so most people should only need to apply once.

If each project needs a full application to explain, you may submit multiple applications, but most of the time this should not be necessary. If there's an idea we like in the "other ideas" section then we'll follow up about it.

whole gale
#

I applied. Fingers crossed our project is interesting enough. šŸ¤ž

sharp rover
#

Love this and it can help me with a current project so perfect timing! Didn’t see this in FAQ: Does the primary applicant (person or company) need to be US based? Being there for the two IRL days is no problem, but I relocated from SF to Aotearoa (New Zealand) recently and my team is distributed. I may even find local folks here to join me in this project. Is that OK?

twin sky
#

There is any section for people looking for a team? Working on local voice model rn

fickle crow
merry ravine
steel frigate
#

@merry ravine if we apply early, can we get feedback to improve the application?

merry ravine
hollow hull
#

Are arts / creation tools welcome in this programme? It seems quite productivity focused

#

Relatedly, is the term ā€˜AI’ used broadly, or does it only mean LLMs and GenAI in practice?

faint magnet
#

Applied today, but I'm planing on bootstraping if I don't get anything

#

Basically I have an OK version of it designed, but I want to take it to the next level

stuck imp
#

if my project (htmx) has nothing to do w/ AI but is focused on improving the core web infrastructure, can I apply?

merry ravine
merry ravine
merry ravine
stuck pawn
fickle crow
rotund ridge
crisp quarry
#

I was so excited to apply after seeing the MEICO cohort last year, but I'm disappointed to find the door seemingly closed to me as my highest ideal is helping people write code, not helping AI write code

#

Is it really the opinion of Mozilla that AI is the only kind of tool that matters? If given the choice between assisting the development of a tool which is a white box (math, programming, computers) and a black box (AI) Mozilla is saying that it believes the most important criteria in developing new tools is that they be black boxes...

#

In other words, would proposals be considered that fulfill the overall objectives: increasing Privacy, Agency, Cost-effectiveness and Reliability of development? Or is the criteria really just "AI tools"...

cosmic laurel
#

so cool

merry ravine
# crisp quarry In other words, would proposals be considered that fulfill the overall objective...

Loved the MIECO program. (If anyone hasn't heard of that, it's this: https://future.mozilla.org/mieco2023/ ) The accelerator is a bit different. One difference is the more narrow theme.

Focusing on a theme helps us put on a better program for participants, because the experts/mentors we work with, credits we procure, and peers we gather are all more likely to be relevant to participants.

So it's not that Local AI is the only kind of tool that matters, but it's an important area where we think we can make the a positive impact right now.

normal mountain
#

Mozilla could make a huge impact for local AI by allocating some more resources to their WebGPU implementation! Lots of people are building interesting local AI tools using wgpu!

slow belfry
rotund ridge
graceful orchid
# crisp quarry Is it really the opinion of Mozilla that AI is the only kind of tool that matter...

hi!

unrelated to the moz org here, and I don’t wish to be abrasive… however… a few bits of context…

this discord’s focus is explicitly on a facet of technological innovation, and moz’s relationship to and with this … exceedingly rapidly growing… incredibly complex and problematic space.

generalizations are easy to craft… and oft misleading

moz has, for decades, been one of the VERY FEW organizations truly focusing their efforts on the human… their rights…. their privacy…. their equal access to and interaction with burgeoning tech..

their mission has consistently been to raise the bar of available and accessible through openness and transparency for everyone..

they’ve walked the walk and earned (at least my) trust… an that’s saying something.

many of the purse strings of companies doing stuff with ai currently are pulled tight when talking about things like privacy, accountability and integrity. there’s a huge short-sighted focus on profit… obvs this is a generalization but it holds more true than not…
AND innovating in this space is absurdly expensive.

so, to me, it comes across as a sour note to suggest with indirect insinuation an exclusive investment in a direction…

given that, and combining it with your chosen handle, i find your question… and tone… unreasonable and frankly kinda insulting… I don’t wish to be hostile or argumentative… but I would ask if you’ve explored the other areas and discords where mozilla is actively stewarding the mission of equality and transparency through investment and initiatives for an opportunity more suited to your chosen QUITE COMMENDABLE method of collaborative and mutual growth?

I’d be willing to bet there’s an avenue more aligned to your path… I don’t know where that might be off the top of my head; but i hope you find it…

crisp quarry
slow belfry
#

as much as you may dislike ai .. its been here since 60 years

#

its just now really more capeable

#

so blackbox or not - it wont go anywhere

#

in my humble opinion - this i dont like ai because we dont understand it - is like saying we cant invest into neuro science because we dont understand the brain

crisp quarry
#

If that's the case, that this event is about the way we've been defining AI for the last 60 years (much of which was not a black box definition) that would seem very different to me than if it's only about the way we've used the word for the last 2 years

slow belfry
#

i used ml for the past 20y profess.

#

mostly vision related - as my background is robotics

crisp quarry
#

At one point Lisp was going to be AI... Now we just call it a programming language

slow belfry
#

and i can assure you we did not understand everything as you think we did

#

there is just little way around ai - yes its overhyped and yes it may plateau - yes we may get a other ai winter

#

but that doesnt mean its not a worthy embarkment

crisp quarry
#

sure, but it also doesn't mean it's the only worthy embarkment, and that's how the industry feels right now...

#

"if you want money, say it's AI"

slow belfry
#

mozilla doesnt just do ai

#

last time i checked

#

you are in the discord of its ai "devision"

#

so ofc here its about ai

#

but thats not all of mozilla

crisp quarry
#

yeah, I really wanted to find the other parts I guess

slow belfry
#

staffers maybe able to help you out there - if there is a discord for it that is

crisp quarry
#

I have one of the more important projects of the AI era I think, but it's purpose isn't to enhance AI's capabilities, it's purpse is to enhance humans' capabilities so that you don't need a language model just to not get an RSI from editing boilerplate code

#

Thats why I don't think I could be funded -- my purpose runs counter -- my purpose is to give people a leg up on AIs

slow belfry
#

i think you are misslead tooling around ai gets good funding .. even if its just dataset related

#

you gotta have a usecase tho

#

so maybe pitch what you have

crisp quarry
#

I believe that, and I believe that I could get funding that way, by making my mission about helping AI instead of making my mission about helping people

slow belfry
#

just tell us what you got instead of beating around the bush

#

how about that

crisp quarry
graceful orchid
crisp quarry
#

I do know, I do that a lot.

slow belfry
#

seems just a bnf style for dom ?

#

at least the CSTML ..

crisp quarry
#

that's not a bad way of putting it, sure -- and the parser core required to write parsers that return these trees

slow belfry
#

right

#

i personaly dont understand why thats in js .. but you will have your reason ..

#

i also dont really see how thats a improvement for humans

crisp quarry
#

It's in JS because it's a UI engine -- I want the raw data structure as close to the HTML DOM layer as it can be so that innovating in UI doesn't mean writing network requests

#

IDEs are currently sort of split into frontend/backend (like the internet) but the narrow waist is LSP, and it's tree-sitter queries

slow belfry
#

well .. sure a lsp is a external dependency for the ide ..

#

but to be frank .. i dont what to have my vim load trees for stuff it doesnt need

slow belfry
#

speed - overhead

#

also updateability

#

lsp's are maintained inpendend

#

if you just push all together you build a cluster f of a monolith

crisp quarry
#

yeeeeeeah!

slow belfry
#

ignore my ignorance but i cant see the benefit of using your stuff over lsp's

#

maybe im just too old that i like tools that work well

crisp quarry
#

What you've said tracks my experience exactly. People believe (somewhere very deep down) that it's possible to build tools that work better. I didn't hear you say that you don't want trees, just that you've learned that you're often more productive without them because of the amount of complexity baked into that implementation

slow belfry
#

if i need an lsp i add the lsp .. i dont want a monolithic monster

#

trees are nice and all but if you work on files 1g+

#

try to load that up in nvim with treesitter

#

good luck

crisp quarry
#

Tree-sitter isn't a streaming parser, BABLR is. You'd see the start of the file with full tree highlighting almost instantaneously, and the overall length of the file would not be a factor in responsiveness (due to incremental rebuilds).

slow belfry
#

just use that as pitch then - get geospacial xml data

#

and show the benefit

crisp quarry
#

The benefit to humans or AI? That's my big question.

slow belfry
#

why one or the other - you had benefits from ai

#

if you ever used a search - you had a benefit from ML / ai

#

specially in the last idk 10 + years

#

its not a mutual exclusive

#

also whats the ai hate ?

crisp quarry
#

I don't think it's hate to say that what should reserve a special kind of love and compassion for people over machines

#

I don't believe in AI trickle-down either. "Make your goal helping AI, and helping people will be a natural secondary effect" is not a very convincing argument I think

slow belfry
#

thats like saying research doesnt help

#

current ai (lets call it llm's) as thats what nowadays mostly is affiliated with it .. arent the bad "tool" here

crisp quarry
#

they are aimed at replacing people though

slow belfry
#

again i like to compare it with neuro science .. should we stop reseaching the brain and try to figure stuff out just because we dont really understand more then then fruitfly ?

#

every change in history had, closed some doors and opend others

crisp quarry
#

And I believe LLMs are here to stay, and that their ultimately replacing people at some tasks is in fact inevitable

slow belfry
#

the human is far from beeing obsolete

#

its just that humans fear change ..

#

same stuff happend when the internet became a thing

#

the industrial revolution

#

.. its just the regular pushback for change

#

but then again im pro ai

#

so i have a bias

crisp quarry
#

Now you're making a clear case though for why as some people's goal is to make money by developing LLMs good enough to replace people. And that I think is the belief that makes people think LLMs are a good investment

#

but it also what makes me want to fight to say "we also need to products whose only purpose is to help people adapt to the tremendous shift in the competitive landscape"

slow belfry
#

thats like investing in faster horses when ford started building cars

#

anyway .. this goes nowhere really

#

if you would have asked what people wanted back then - it would have been the faster horse

#

^^

#

want and need / whats going to happen are often 2 different paths

crisp quarry
#

I've used exactly that Henry Ford quote several times now in reference to why I'm building something so at odds with the zeitgeist on where the technology of coding is going

lyric solstice
exotic python
merry ravine
merry ravine
lyric solstice
merry ravine
keen pumice
#

who's applying to this one @graceful orchid you in ? wanna link up ? if i have a program question , can i straight up pitch my idea in this channel ? @graceful orchid according to you is that a good idea ? i dont wanna shoot meself in the leg either šŸ™‚

graceful orchid
keen pumice
#

far out , thanks a ton šŸ™šŸ»

rigid wedge
#

hi all, in the application form are you allowed to use Markdown or should it be only plain text?

merry ravine
rigid wedge
keen pumice
magic ridge
#

Hello everyone! We at Awana Digital are excited to propose using local offline-first AI to enhance accessibility for our indigenous partner communities, who are earth defenders across the globe. Our platform, the Earth Defenders Toolkit, aims to empower these groups, but current text-based formats are not suitable for our oral-first audience. By integrating voice-input/output into a conversational chat interface, we believe we can make these tools truly accessible while keeping sensitive field-collected data secure and local. Given our team's ongoing commitments to field training and UX feedback collection for projects like Mapeo, Terrastories, and the Earth Defenders Toolkit, we would like to ask about the possibility of being flexible to rotate team members participating in the weekly Mozilla Builders Accelerator program. Thanks! Our website: https://digital-democracy.org/

earnest estuary
fickle crow
fickle crow
earnest estuary
keen pumice
lyric solstice
#

Hey, sorry to bother, but for the application what kind of billable items should go on the list. Since the AI is local and won't really acrew any cost, so what other factors are you guys considering?

whole gale
# lyric solstice Hey, sorry to bother, but for the application what kind of billable items should...

On our application, we split the 100K up like so:

• $40,000 for continued Software Development: Enhancing AI algorithms and integrating new features based on pilot feedback.
• $20,000 for Pilot Testing and Evaluation: Expanding our pilot programs to more XXXX, including costs associated with onboarding, training, and data analysis.
• $15,000 for Security and Compliance: Ensuring that XXXX adheres to the highest standards of data protection and privacy laws.
• $25,000 for Outreach and Education: Developing training materials and conducting workshops to educate XXXX on the ethical and effective use of XXXX.
This investment will enable us to refine XXXX, expand its reach, and ultimately provide XXXX with a powerful resource that XXXX.

Hope that helps. šŸ‘

dense drift
#

Hello! I sent my application but I didn't receive an email confirmation. Is there any chance that you confirm that you got it šŸ™

exotic python
magic ridge
fickle crow
fickle crow
fickle crow
fickle crow
magic ridge
keen pumice
#

great to see the ideas trickling in šŸ™‚

devout anchor
#

Are there any word or character limits for any of the answers?

#

How would you answer "Project Type" for a company that has an existing non-AI-related product that wants to create a new Local-AI-related project that will ultimately be useful for the original product?

wicked oxide
#

oh woah i totally missed thi

fickle crow
fickle crow
lilac quartz
#

@fickle crow Could you clarify what type of "non-dilutive funding" you are planning to use? A simple grant? Some debt or revenue based financing? I couldn't find a clear explanation in the FAQ. Thanks!

fickle crow
devout anchor
#

@fickle crow I have some time right now, but if you're busy, we can schedule for a time that works for you. Would love to be able to refine the idea with you, thanks for your offer. Would Friday afternoon work for you?

old swallow
proud arch
#

This is all that comes up when I click on the "Apply+" button.
Where's the actual application?!?

"Welcome to the Mozilla Builders Accelerator Application for 2024.

We encourage you to submit your application as soon as you’re ready to apply. We read applications on a rolling basis and applications that are submitted early may be selected for an early interview.
People who apply by the early application deadline will hear back by August 1st. Applicants that apply by August 1st will be notified by August 31st 2024 or earlier.
If your application is promising, we will invite you to interview with us. Most interviews will be held by video conference from July to August.

Funding amounts are determined by the size of a project and team. Ultimately we are interested in ensuring that there is sufficient funding to successfully complete their project by the end of the accelerator. Be sure to be clear about the scope and deliverables of your project in your application.

Questions about your application? Join the Discussion Thread in the Mozilla AI Discord.
Need an invite? Join Mozilla AI on Discord

We can't wait to learn more about your amazing project!"

jaunty summit
fickle crow
# wicked oxide unfortunately it seems not

maybe a notification setting to tweak Notifications Settings --> check to see if role mentions are suppressed and mentions are ok if you would those builder pings. We will ping that role with any new opportunities for builders. šŸ™‚

fickle crow
#

I would suggest applying! I think as long as there is a local device component you are within the theme. Maybe you rely on the local network/internet for some features but its also capable of running in some capacity on the device offline for example.

See also John's responses in thread: #1245083732319408195 message

uncut cargo
#

I have a question about the application form:

For the Project Type selection, at what point in time does this apply?

When I started my project, it was a proof of concept / new product. By the time the program starts, I'll be adding more features to an existing project.

proud arch
devout anchor
#

@fickle crow we're hoping to submit our application by the early deadline. Is it by July 8th 11:59pm? Any particular timezone?

green ember
#

Sorry, if this was already asked before: But can you give any indication on how many applications you received so far?

pine pine
#

@fickle crow @merry ravine - Truly sorry to even ask, however we had to validate a few additional facts regarding participation prior to submitting, and we weren't able to get it submitted until today. Is there anything we can do or anyone we can ask about inclusion for the early deadline evaluation? Please advise, šŸ™ Thank you in advance! (A DM would suffice if you prefer.)

fickle crow
# pine pine <@376601692189360138> <@177642687934103553> - Truly sorry to even ask, however ...

Thanks so much for your interest and time with the application. Although you missed the deadline (#announcements message), it's great you got your application in at this point. As per John's previous message:

Early applicants have the advantage that we'll review their applications sooner, we'll be in touch longer, they have more time to refine the ideas, etc. But we won't be able to give specific, critical feedback to every applicant. We'll reach out to applicants who pass the initial review to ask questions and set up an interview.

Early or not, we have a small team of reviewers and we appreciate your understanding there. cc: @devout anchor too in case you missed it.
@green ember I'm not aware of the current total app number as I'm not a reviewer. Thanks all for the questions.

pine pine
devout anchor
frail moth
#

Hi @fickle crow! Are there any restrictions on the types of OSS licenses that are allowed for proposed projects?

oblique cedar
#

when will we start hearing about the early consideration folks? šŸ™‚

fickle crow
fickle crow
runic summit
#

FAQ kind of touches on this ("Complete your project by the end of the program"), but just to clarify: requested budget should be just for 12 weeks of work, not for e.g. a whole year, perhaps using the 12 weeks to get to launch?

oblique cedar
fickle crow
fickle crow
frail moth
devout anchor
#

Is there a set number of applicants/teams expected to be in the accelerator program?

mild sparrow
#

We just submitted our application for local first Agent development and testing. šŸ¤žšŸ˜…
Good luck to everyone! If we get through, it would be awesome to meet some of you in person.

devout anchor
# devout anchor Is there a set number of applicants/teams expected to be in the accelerator prog...

@fickle crow I believe we made the early application cutoff (at least according to our own timezone) but I’m wondering if there’s any way to get confirmation that our application is being reviewed as part of the early application process or not? We haven’t received anything since applying and it would be great to get a ā€œyour application is being processedā€ sort of message if possible.

fickle crow
fickle crow
fickle crow
hexed musk
#

If I've made decent progress on my project, is there a way to amend my application to expand the scope (and also reduce funding amount since I have a better idea of what it takes to get it done)

fresh flower
#

Hi @fickle crow

Questions about the application: I know the expectation is that successful applicants will 'work on your project full-time during the 12-week program', an equilivent of 450hrs. I wanted to check that - as a nonprofit - we could apply with a plan to collectively spend that amount of time (450hrs) on the project, ie it might not be 'one FT member of staff working solely on the solution for 12 weeks'. Would we still be eligible? We would have a designated point of contact for engagement in the community/participate in in-person events etc.

Thank you for your guidance!
D

fickle crow
fickle crow
fresh flower
peak field
#

@fickle crow
My name is Jonas, CEO of EclipseSource. We build open-source development tools and AI integrations. We are huge fans of LLamaFile and recently found out about the Mozilla Builders program.
We are currently already working on a project that we believe is a pretty good match with your general vision. However, as we have already started on it, we are maybe not perfectly in line with the funding program. I still wanted to share the idea with you and see whether there is any interest in alignment, collaboration or exploring synergies.

The Project (Theia AI IDE):
We are currently developing an open, flexible and transparent AI-driven IDE (based on the Theia IDE). The core principles include a strong emphasis on privacy, where developers have complete visibility and control over data exchanged with AI models. Specifically, it allows users to connect to local LLMs, enhancing privacy and security. We integrate LLamaFile for this. Other features include reviewable and adaptable prompt templates as well as a transparent local history of the full communication with the underlying AI.
Our goal is to ensure that developers can customize their IDE's AI interactions, offering a secure (local) and adaptable development experience.

If you find this project interesting or want to learn more, please let me know, maybe we can find a way of collaborating. Looking at the Builders program and watching your presentation, I believe we share a vision and see similar issues in the market.
I am also happy to submit a formal application although as mentioned, we are already very actively working on this and will likely continue to do so for quite a while.

fickle crow
old swallow
jaunty summit
#

Looks like some automated rejections went out! I was part of two proposals, one early bird, one later, and we received both rejections this afternoon. How about a thread of things that didn't make it, to see what each other is working on?

torn patio
#

Hi! Thanks for your hard work reviewing all the applications! Are all decision emails sent for early applicants? I applied before early deadline but have not gotten any emails yet. Thank you!😊

hexed musk
old swallow
#

Good luck everyone who's still waiting!

#

Mozilla's been a terrific company for me to work with. I'm glad others will have that opportunity too. For those who don't make it in, I'm sure there'll be other opportunities to try in the future.

hexed musk
#

I applied because I randomly stumbled on your tweet @old swallow!

fickle crow
fickle crow
#

@fervent crag Applications close in about 12 hours (midnight PT), thanks for checking

languid tiger
#

I am curious. What are the requirements for success, and how long should I wait to receive the rejection or acceptance email as an early applicant?

How many projects does Mozilla plan to fund? If a project is selected for the first phase, what are the chances of it being rejected in the second phase?

Still I have 100+ more questions

misty osprey
#

I have submitted my application for the program. Thanks team for all the resources and faqs
Good luck everyone

sharp rover
#

I’ve never been more nervous applying for anything! And of course left it to the last day thus ensuring high stress levels for the past two weeks. Good luck everyone and fingers crossed get to meet some of you soon. Clicking the submit button in 20 mins and counting… it’s like a rocket launch 😬

sharp rover
fickle crow
#

I've had a few questions about when the accelerator selection will be publicly announced. We can expect that sometime in September. šŸ™‚

sharp rover
#

As with others in here pretty gutted we didn’t get selected and have been moping. But it was a really productive exercise to go through and I wanted to share our deck in case it sparks anything with anyone. The R&D work we were seeking to find helps scientists with storytelling collection in the field, then curation on our (existing) platform. So: science communication.

Because Mozilla can’t give out individual feedback I’m open to any thoughts this community may have!! TIA and massive congratulations to the successful parties. I can’t wait to see which submissions were greenlit and follow their journey. Kudos to Mozilla and the team behind Builders for this initiative and commitment to AI the ethical way šŸ‘

https://docs.google.com/presentation/d/e/2PACX-1vQPsHy8kFeXfGFcyF2T5L996TyrhLERj4M02y1xUQ6Xcg7CYr1r3yeKZbqRdtgI-Qgm-7FEvpzISon-/pub?start=false&loop=false&delayms=3000

(You may need to open in external browser)

fickle crow
exotic python
#

All I think is that mozilla AI was maybe not the right accelerator? I think they're mostly looking at tools and models that act as the building blocks of open and collaborative AI

hollow mortar
#

I would love to definitely be a part of this.

fickle crow
frank river
#

Will there be a 2025 cohort?

We have started an AI which detects and prevents suicide, homicide, terrorism, sexual trauma, and 27 different crises by getting people involved before it's too late. After the lockdown at my 5 year olds school I Am more gung ho than ever. Linkedin Live event Demo, CTA and AMA set up.

Join our first event this morning at 0930 PST if you like. Have a look at the guest list. Pretty good for networking as well as learning about the future of dealing with the problem instead of the symptom. If not me then who? I go for the throat when I take down a problem.

https://www.linkedin.com/events/7256767917122375680/comments/

We have also been accepted into the Founders Institute, however Mozilla has allowed me to teach myself how to code over the years by studying its docs, learning XUL and the old hack competitions back during the browser wars. Not to mention once again pulling me into the forefront with llama.cpp last year.

fickle crow
#

No dates planed for the next builders cohort yet. Thank you for sharing more about your project and the event!

fickle crow
#
Mozilla Builders

Mozilla Builders works with open-source AI developers in two ways — we Collaborate and we Accelerate. Collaborations are bespoke co-development and sponsorship agreements to advance SOTA open-source AI. The Accelerator funds and supports work vital to the open-source AI ecosystem. It provides $100,000 in non-dilutive funding and a focused 12-wee...

fickle crow
primal kite
#

any news on 2025 programs, if any?