🤙🏽🌺 Pau Hana Time fo’ Da Big Tech A.I.: Local Style!

⬇️ Pidgin | ⬇️ ⬇️ English

Da non-profit Allen Institute fo’ AI, lead by one top-notch computer scientist who wen’ sell his company to Apple, tryna make everybody get chance fo’ access da latest kine research. Ali Farhadi no stay any kine tech kind troublemakah.

Da buggah, 42 years old, stay one solid computer scientist, teach at University of Washington, and he even wen’ start one company Apple wen’ buy. He wen’ work fo’ dem till like four months ago. 🍎

But dis Farhadi guy, aftah he wen’ become da big boss of da Allen Institute fo’ AI in July, he been talking stink about needing “radical openness” fo’ make sure everybody get chance fo’ do research on top da new kine A.I. Everybody tink dis da most major tech tink come out in long time. 🤖💡

Da Allen Institute wen’ start dis big kahuna plan fo’ make one free A.I. dat can compete with da big boys like Google and even da fresh blood like OpenAI. Da way dey like do um is called open source, where all da uddah smart heads can poke around da system and da info inside. 🔍📊

Da way da Allen Institute stay doing tings, puts dem right in da middle of dis big scrap about if da new A.I. should be open or stay secret. If everybody get access to da new kine A.I., like da ones powering chatbots from OpenAI and Google, will dat mean mo’ cool stuff goin’ come out? Or goin’ cause plenny palaka online?

When peeps talk about “open” fo’ da new A.I., get choke ways fo’ explain. Old school kine software, peeps could see all da ins and outs. Can see if get buggas, can suggest changes. Get rules fo’ decide if goin’ change um. Dat’s how Linux, Apache, and Firefox stay workin’. 🌐🛠️

But dis new A.I. kine mo’ tricky. Needs plenny data fo’ train and adjust da system. Even if get good intentions, da path Allen Institute taking kinda sketch.

One smart dude from Harvard, Aviv Ovadya, say decisions about if da A.I. stay open or not super important, and we need errybody from around da world fo’ decide wat kine tech should be out dea fo’ everyone. 🌍🔐

Dis generative A.I. super powerful, but also kinda wild. Can write emails, poems, and even school work real fast. But also can act all kookoo and make up stuffs. Da big companies like OpenAI and Google, dey keep deir new stuffs on da DL. Google, especially, used fo’ share all deir tings before, but now, especially with deir A.I. Bard, dey acting all stingy. 🤫

Dey say if dey let everybody see inside, den bad guys might use da tech fo’ spread fake news and scams, or even worse stuffs. 🕵️‍♂️🚫

Peeps who like da open system know get risks, but tink mo’ smart guys working on top da problem da bettah way fo’ handle.

Meta wen’ come out with one A.I. model called LLaMA dis year and had errybody talking. Farhadi tink dey doing good, but not enough. He like show everybody all da secrets of da A.I., from da start to da finish. 📖🔍

Da Allen Institute took da first step, releasing all kine data fo’ train da A.I. models. Get stuffs from online, books, school tings, and computer stuffs. Dey make sure take out all da personal info and bad talk. 📚🖥️

Dey got da biggest set of open data right now, Farhadi said. Since August, been downloaded ova 500,000 times from Hugging Face, one website fo’ open-source A.I. tings. 🤗📈

At da Allen Institute, dey goin’ use da data fo’ make one big A.I. program called OLMo, coming out soon. Farhadi call da uddah big A.I. models “black box” but he pushing fo’ “glass box”, so can see inside and talk story about how stay work. 📦🔍

Only get few big A.I. models like wat Allen Institute tinkin’ about. Include Meta’s LLaMA and Falcon, one project Abu Dhabi supporting.

Da Allen Institute perfect fo’ dis kine big A.I. project. “Get plenny money but stay true to da school, and been helping advance open kine science and A.I. stuffs,” said one computer scientist from Carnegie Mellon University.

Da Allen Institute no stay alone, teaming up with uddahs. Dis year, da non-profit Mozilla Foundation gave $30 million to one new company, Mozilla.ai, fo’ make open-source software fo’ help with open A.I. engines, like da one from Allen Institute. 🤝💸

Da Mozilla Foundation, started in 2003, no like see only few companies controlling all da tech power. “Only few guys, all from da West Coast, trying fo’ take control of da A.I. before even take off,” said da big boss of da foundation, Mark Surman.

Farhadi and his crew trying fo’ make sure dey control da risks. Working on ways fo’ see if da model acting right and stop um from doing bad stuffs, like being racist or helping make dangerous stuffs.

Farhadi tink da safety nets in da big chatbot models easy fo’ smart hackers fo’ break. “My tink is, should no let dat kine info stay inside,” he said.

Some peeps goin’ do bad stuffs with dis tech, Farhadi said, just like with any powerful tech. Da job fo’ everybody, he said, is fo’ understand and control da risks. Being open, he tink, is da best way fo’ make sure stay safe and share da money opportunities.

“Rules alone no goin’ work,” Farhadi said.

Da challenge fo’ da Allen Institute big. Need choke computer power fo’ make and adjust one big A.I. model. Farhadi and his guys tink get bettah ways coming. But he tink da Allen Institute project goin’ cost $1 billion in computer work next few years. He trying fo’ find support from government, private companies, and rich tech guys. But no like talk who helping.

If he make um, da real test is fo’ make sure get solid support fo’ da project.

“Need choke open-minded peeps fo’ really challenge da big guys,” said Surman from Mozilla Foundation. “And da real challenge is just staying strong and neva giving up.” 🤜🏽🤛🏽🌈

---

NOW IN ENGLISH

🤙🌺🏄 Brahdas Shaka A.I.: Going Head to Head Wit Big Tech!

No mistake, Ali Farhadi no like make waves. He stay 42, solid in da computer science world, even get choke respect teaching at University of Washington. Had one company dat Apple wen buy and he was chilling wit dem till four months ago. 🍏👨‍🏫

But now, Farhadi taking da big chair at Allen Institute for AI and shouting out, “We need mo’ aloha in A.I., let everybody in!” 🎙️🤖

He and his gang at Allen Institute get one master plan: Make one free A.I. dat can throw down with the big boys like Google and even da rising stars like OpenAI. It’s like when your uncle lets everybody sample his secret BBQ sauce at da family luau. The goal? Make everything open so da tech ohana can check ’em out and even add their own mana’o. 🌴🔧

This new vibe from Allen Institute, which stay in Seattle, making plenny people talk story. How open should this next-level A.I. be? Like sharing your auntie’s malasada recipe or keeping it stashed away? 🍩💭

Usually, when peeps talk about open-source, they mean let everybody look under da hood, see the mechanics, and even suggest improvements. It’s like everyone helping fix the family van. 🚐💡

But this new A.I., ho, it’s more than just seeing the gears turn. Training these A.I. models is like teaching your keiki – takes time, love, and plenny data! 📊👶

Even though the intent is good, experts giving da shaka sign 🤙 with caution. “The choices we make about A.I. and openness, brah, big kine impact for everyone!” says Aviv Ovadya from Harvard. He’s thinking maybe the world gotta come together and decide what should and shouldn’t be released. 🌏🔓

Generative A.I., while being the bomb, sometimes act all wonky. It can write emails, poetry, and even school essays like one pro, but sometimes, it gets all lolo and imagines stuff that’s not real. Kinda like your cuz after too many beers. 🍻🤪

The big names in chatbots, like OpenAI (backed by Microsoft) and Google, they keeping their cards close. Especially Google, who used to share all kine things, but now, they acting all secretive with their new tech, Bard. Their reason? They no like the bad guys get the tech and make all kine rubbish online. 🕵️‍♂️💻

But the open-source supporters, they say more brains better than few. More heads to think, more hands to help. When Meta wen bring out their new A.I., LLaMA, had big noise. Farhadi gave props, but thought they could do even more. 🚀🐐

Farhadi’s vision? Share everything – the tech details, data, adjustments – all of it. Allen Institute wen take the first step, releasing one massive collection of data. No personal info, no bad words, all clean. Kinda like your grandma’s old photos. 📷👵

The data from Allen Institute, biggest out there right now. Since dropping it in August, over 500,000 downloads on Hugging Face, the go-to for open-source A.I. goodies. 🌐💾

Farhadi calls most A.I. out there “black boxes”. But he’s pushing for “glass boxes”. Like, “Open everything up, then let’s talk story”. 🖤🔍

Only few of these massive A.I. models are out there for all to see. Allen Institute though, they in a good spot to make it big. Got the funds, the academic street cred, and one history of pushing for open tech. Zachary Lipton from Carnegie Mellon says they’re on the right track. 🚂🧠

And they no going solo. Teaming up with others, like the nonprofit Mozilla Foundation. These guys just put $30 million into Mozilla.ai to make even more open-source tools. They worried about only a few big techs controlling everything. 🕸️💰

Farhadi and his crew, they not blind to the risks. They working hard to make sure their open strategy no backfire. Making sure the A.I. behaves and doesn’t do bad stuff, kinda like putting up fences at the beach so no one gets hurt. 🏖️🚫

“People will misuse this tech,” Farhadi says. But the challenge for all of us? Understand and manage the risks. His belief? Keeping things open, the best way to find safety and share the opportunities. 🛡️🌅

It’s not gonna be one easy ride, though. Need massive computing power to build and refine these big A.I. models. Farhadi and his ohana, they’re looking for help from government, private sectors, and tech angels. But, he’s keeping mum on who’s backing them up. 🤐💼

If he makes it, the next big test: Building one solid community to keep the project alive and kicking.

“Need one big ohana of open supporters to make waves against the big kahuna,” says Mr. Surman from Mozilla Foundation. “And what’s crucial? Just gotta have patience and stay pono.” 🌊🤙