:format(jpg "Is CryptoAI truly a match made in heaven?")
Trent McConaughey's father was a pig farmer in rural Canada. Not much entertainment. As a child, in the 1980s, his father bought him a computer with reference books. Trent couldn't play video games at the arcade, so he taught himself to code and created his own version of Pac-Man. When he was about 10 years old, he found a used book at a flea market.
The book is dedicated to artificial intelligence.
“I’ve been passionate about it my whole life ever since,” McConaughey said. In the 1990s, he downloaded neural networks from “bulletin boards” before the advent of the Internet; pioneering research into the use of artificial intelligence for printed circuit board design; and worked for artificial intelligence startups until he “discovered the blockchain bug in 2013.”
While many in the cryptocurrency industry have only recently gotten into AI, McConaughey has been doing it since the beginning. “Blockchain has been a priority of mine for several years,” he wrote in 2017. “But AI has always been my first love, and I never let it go.”
McConaughey's Ocean Protocol project is one of a growing number of cryptocurrency AI startups looking to use blockchain to improve artificial intelligence. But this is where things get confusing. What does “crypto meets artificial intelligence” mean? Since the ChatGPT hype began, many Web3 users, including myself, have written about how cryptography and artificial intelligence can fit together. But often these concepts are vague and abstract, glorified in the vague sense of “chocolate and peanut butter combined.”
Perhaps the important point is that you typically see expectations of "Web3 and AI need each other" in the Web3 space and rarely in the AI space, which creates a sense of one-sided university competition. Even many web3/AI supporters are unhappy with Opium. Mrinal Manohar, CEO of Casper Labs, which works on solutions that combine blockchain and artificial intelligence, said: “I'm actually very disappointed with the story that's unfolding. “There’s this feeling of: Let’s throw some blockchain fairy dust in there and everything will be fine.” This is not true.
So, the goal here is to dig deeper and, more specifically, explore how Web3 can help develop safe and ethical AI. There are dozens of projects with applications ranging from counterfeit detection to improving AI security (Alison Deutman provides an excellent overview), but here we'll focus on two key areas: data and computing.
Any AI model is only as good as its data. And training these AI models requires so much computing power (and energy) that Bitcoin mining looks like a pocket calculator.
So how can Web3 help?
Let's start with his data. Anyone who has tried ChatGPT knows that the AI has a tendency to “hallucinate,” which is a good euphemism for “bullshit.” I like to think of ChatGPT as an employee who is so eager to please you, so hungry for your affection, that he lies to make you think he's good at his job.
The problem of hallucinations is complex. There is no clear answer. Once data is extracted from the LLM, it is essentially a black box, and even the smartest engineers in the industry cannot accurately determine which inputs cause which outputs. (Here I present a brilliantly written academic article that ends with a shrug emoji.) Additionally, based on recent cases by authors such as Sarah Silverman and Michael Chabon, we have seen that intellectual property based on copyrighted works has been created illegally. . .
“Blockchain can eliminate all of this,” Manohar said. “Blockchain is the most powerful copy protection technology in the world.” Blockchain, with the ability to theoretically track data for each block in the chain (with more emphasis on “theoretically”), allows data to be checked for copyright infringement, bias, or inaccuracy.
The use of blockchain can not only control the data used to run models, but also expand the pool of potentially useful data. Think about healthcare or financial information. Given real privacy concerns, “those are two data sources we can't touch,” says Duetman, CEO of the Forecasting Institute. “Big data models cannot achieve this.”
But what if people are interested in providing their personal information, and this can be done in a secure way? Ali Yahya, general partner at venture capital firm Andreessen Horowitz, calls this data the "long tail" and says Web3 is the only way to take advantage of it. “There is data in this world that doesn’t exist without long-term data,” Yahya said.
Yahya explained that AI teams traditionally create data models, a top-down structure where you explore the world and find data, much like Waymo puts cameras in cars to film every road on the planet. “But probably the best way to do it is bottom-up,” says Yahya, “with ordinary people providing data from their own mentors.”
But these ordinary people won't do it for free. They need incentives, which brings us back to Trent McConaughey. In 2017, he envisioned Ocean Protocol as a data marketplace capable of “connecting those without data” that would “propagate the network effects of AI data, thereby distributing power and resources.” .
In the next sentence, McConaughey added almost matter-of-factly, “But technology isn’t easy.” The challenge: "You have to store data about who owns it with strict user and privacy controls," McConaughey explained. “There needs to be agreement with governments and regulators on privacy and data sharing. It needs to be decentralized. It needs to be a big project, not just some shiny toy technology. Decentralizing technology on a large scale is very difficult.”
Read also: Alison Duetman - How cryptocurrencies can help protect artificial intelligence
That is why it took five years to create the maritime market infrastructure. (Another OG AI project, SingularityNET, has spent the last five years building a decentralized AI marketplace.) Sometimes it doesn't even feel like an AI-focused project. For example, a recent blog post states that the next step in ocean development will be “value cycle management” and that “the most promising and mature area is decentralized finance.”
The point of all these little things is that before chocolate and peanut butter crypto-AI becomes a reality, a lot of pipes need to be installed. McConaughey has a long-standing passion for artificial intelligence and has been developing it for many years, but he admits that "we haven't been able to create a lot of artificial intelligence products out of the box."
The situation may soon change. He said the market for data now exists, the incentives are there, and “we can now experiment a lot more with AI.”
Difficult to calculate, difficult to verify
An AI game requires computing, and computing requires computer chips like the Nvidia H100, and those chips are in short supply. (OpenAI even announced that it was delaying training its next model due to a chip shortage.)
While this shortage may only be a short-term crisis, it is likely to continue—or get worse—as interest in data processing grows. What happens when OpenAI trains ChatGPT 7? Could demand for chips exceed supply?
“We don’t have enough space to build large data centers,” said Ben Fielding, co-founder of Jensyn, another project combining cryptography and artificial intelligence. “You hear about going to Antarctica [to build data centers]. It gets a lot more complicated." That's why Gensyn is developing a bold plan to expand its global supply. “You don't just have one big data center,” Fielding said. “Now you have every data center on the planet.”
Genshin's idea: users will be rewarded (with some kind of tokens) for providing the network with their idle computing power, and it will become a decentralized system capable of training artificial intelligence. There really are precedents for such cryptographic projects. FileCoin encourages users to provide unused data storage in the cloud; Helium rewards people who provide local Wi-Fi. Both projects began with launches to the Moon and are now both successful.
Read also: Chris Call-Beswick – Artificial intelligence is killing venture capital interest in cryptocurrencies
Genshin has a better chance.
The team faced a fundamental problem that made what it did difficult: verification. In some ways, their job is more complex than securing the Bitcoin network. “Bitcoins are difficult to calculate, but easy to verify,” said Harry Grieve, another co-founder of Gensyn, explaining that while proof of work requires a lot of computing power, the calculation itself is relatively simple. “Machine learning is hard to compute and hard to test.”
The problem concerns the age-old problem of “mistrust”. If you send me bitcoins, you don't know me and you can't trust me, so the genius of the network is that it verifies both sides of the transaction. But it's fixed. If you're actually processing the data that will train the AI, like images or recordings, it's actually a moving target that needs to be checked against other reference points in the network and everything needs to happen down the line.
Then things get more complicated. The flexible nature of neural networks means that the entire system is always dynamic. "You can't just take two parts and test them separately," Fielding said. The details can be overwhelming, but the basic idea is that each piece of the puzzle needs to be compared to the whole puzzle, which is in constant motion, and this needs to be done using a kind of zero knowledge. Try to maintain confidentiality.
In addition, there are "distributed system problems". Let's say you, me and Taylor Swift offer our computers to train an AI model. Sounds good in theory. But in such a decentralized system, who decides what part of the work your idle computer does, what part my computer does, and what part Taylor Swift does when she's on stage? Such an infrastructure must be created. “It’s a difficult problem,” says Yahya (a16z – Gensyn investor), “but it’s not an insurmountable problem.”
In contrast, centralized systems are relatively simple. Easily manage resources. Verification provided. You will not experience delayed headaches. This is why centralized solutions like OpenAI and DeepMind are getting better. The proposition of Gensyn, Ocean and other similar projects is that yes, what they do is difficult and slow (for now) and it will take time to achieve it, but the end result will be artificial intelligence free of powerful technologies. It's worth waiting for this to happen.
On the other hand... does this decentralized solution need a blockchain?
Need a blockchain?
I asked Yahya directly. If real privacy is cryptography, yes, but not necessarily cryptocurrency?
It's easy for him. “Cryptography alone cannot create this incentive zone,” Yahya said. “It cannot create a decentralized asset market the way blockchain can.”
This may be true, but it would be an exaggeration to confuse all the benefits of cryptocurrencies with the benefits of cryptocurrencies themselves. “You don’t need blockchain for anything related to cryptography,” Duetman said, noting that much of the AI field is focused on cryptographic solutions that have nothing to do with Web3. And while it's true that decentralized artificial intelligence is a goal that many support, it's also true that many open source solutions—like the promising work of Llama 2—seem indifferent to tokens or cryptocurrencies.
It's also possible that Web3 is not only unnecessary for ethical AI development, but may even be harmful. Illustration - AI এরসবচেয়ে খারাপ ভয়ের সাথেএ কত্রিত করুন৷ DAO দ্বারা চালিত একটি AI যেটি একধর Ultron, Ultron
- ভিত্তিক AI-এর জন্য সেরা পরিস্থিতিতেও, বিশেষজ্ঞরা নয়। পরিকাঠামো তৈরিতে সময় লাগবে।
"Advice" ানের জন্য চাপ দিলাম। "Read more "Performance" All rights reserved. """""" "
আরও পড়ুন: জেফ উইলসার - 10 উপায় ক্রিপ্টোগ্রাফি → পারে (বা আরও খারাপ হতে পারে)
এটাও অস্পষ্ট যে কিভাবে ডেটা সলিউশন মত) শেষ পর্যন্ত কম্পিউটিং সলিউশনের সাথে যুক্ত হবে (যেমন জেনসিন)। → User interface, user experience এবং ব্যবহারে সহজ হতে হবে। """"""" " "Advertisement" Illustration "
Pure UX Web3 Web3 ত সমাধান কাজ করে। → → র দেওয়া যাক।
প্রকৃতি 2.0
রিম বুদ্ধিমত্তার প্রতি আচ্ছন্ন। → """""" ে? েছিলেন। AGI কিন্তু করতে সামান্য-বা সম্ভবত কিছুই করা যাবে না। AGI আমাদের মাংসের ব্যাগগুলি All rights reserved.
পিঁপড়ার কথা বিবেচনা করুন।
McConaughey ত্র প্রশ্ন হল কখন। য AGI হল মানুষ এবং আমরা পিঁপড়া। “Advertisement” পারে এবং বলতে পারে: “আরে, আমরা চাই আপনি মানুষের ম তো স্মার্ট হওয়া বন্ধ করুন। দয়া করে পিঁপড়ার মতো স্মার্ট হোন।" মানুষ 100 গু Please contact us ন, এবং "এটি বেশ স্পষ্ট যে AI মানুষের চেয়ে 100 গুণ বেশ "
? আমরা শুধু আমাদের বুট অধীনে তাদের পদদলিত করব.
যদি আমরা এই ভিত্তিটি গ্রহণ করি, তাহলে যৌক্তিক Illustration প্রথমটি হল অনেক লোক যা নিয়ে কথা বলছে: A.I. মমর্থন All rights reserved. লিকে "আলগা করা"৷
Illustration ন। (?) তে হবে, তবে সম্ভবত আমরা হেরে যাব।)
তাই তিনি দ্বিতীয় সমাধান সম্পর্কে আশাবাদী, যেটির বিষয়ে প্রায় কেউই কথা বলে না এবং যেখানে জিনিসগুলি আরও আকর্ষণীয় হয়ে ওঠে: কৃত্রিম বুদ্ধিমত্তা ব্যবহার করে মানুষকে "সুপার ইন্টেলিজেন্ট" হতে এবং আমাদের মস্তিষ্কে কার্যকারিতা যোগ করার অনুমতি দেয়। তার জন্য, এটি "প্রকৃতি 2.0" এবং মানব বিবর্তনের পরবর্তী পর্যায়। এটি মানবতার একটি ভবিষ্যত যেখানে, ম্যাককনাগেই বলেছেন, "আমরা আমাদের রক্তে আবদ্ধ নই।"
সঠিক সমাধানটি এখনও নির্ধারণ করা হয়নি, তবে সম্ভবত আমাদের মস্তিষ্ক স্ক্যান করা হয়েছে এবং একটি কম্পিউটারে আপলোড করা হয়েছে, অথবা সম্ভবত আমাদের সমগ্র দেহগুলি পরমাণু দ্বারা পরমাণু পুনরুত্পাদিত হয়েছে। শেষ পর্যন্ত, ম্যাককনাঘি বিশ্বাস করেন যে আপনার বয়স যখন 90 বছর এবং আপনার শরীর দুর্বল হয়ে যাচ্ছে, তখন হয়ত আপনার নতুন সাইবার্গ স্বয়ং বুঝতে পারে যে "মাংসের ব্যাগটি পথে আছে" এবং আপনি "এটি পেরেকের মতো কেটে ফেলেছেন।" (তিনি একটি 2016 প্রবন্ধে এই ধারণাটির রূপরেখা দিয়েছেন, "এআইয়ের অস্তিত্বের হুমকি: একটি পুনরুদ্ধারকারী বায়ো-নার্সিসিস্টের প্রতিফলন।")
তারপরে তিনি একটি আকর্ষণীয় ধাপে ধাপে দৃশ্যকল্প অফার করেন যা আমি মনে করি অনেক অর্থবহ।
Понятно, что все это на данный момент всего лишь научная фантастика. МакКонахи признал, что до появления этой технологии еще годы или даже десятилетия. Но затем он предлагает интересный пошаговый сценарий, который, на мой взгляд, имеет большой смысл.
Все начинается с чего-то очень распространенного в современном мире: представьте, что вы носите наушники Apple и носите их с собой весь день. Вы общаетесь с друзьями (отправляете и получаете сообщения) и общаетесь с помощью последней версии Siri. Это не надуманно.
А теперь представьте, что вы делаете то же самое, но вместо AirPods вы носите очки с простым ЖК-экраном — полную версию никогда не популярных очков Google. Эти очки позволяют отправлять и получать сообщения, взаимодействовать с чат-ботами с искусственным интеллектом и выполнять запросы в Интернете.
Возможно, в очках есть функция отслеживания взгляда, поэтому вы видите инструкции на экране так, как будто щелкаете мышкой. Опять же, это не надуманно, потому что именно так работают новые очки Vision Pro от Apple.
Теперь представьте себе поворотный момент. Эти очки оснащены функцией «сканирования мозга», что означает, что вы можете нажимать на ссылки, думая, а не нажимая кнопки или даже глядя на них. Это не так странно, как кажется. «На данный момент это довольно практично», — сказал МакКонахи. Например, интерфейс «мозг-компьютер» недавно был использован, чтобы позволить парализованной женщине, которая не имела физического контроля над своими голосовыми связками, снова говорить с помощью сигналов мозга. Одна из примерно 347 компаний Илона Маска, Neurallink, спокойно работает над решением такого типа мозгового интерфейса.
Следующий поворот: эти очки постоянно записывают все вокруг вас, и да, это буквально сюжет эпизода «Черного зеркала». Очки сохраняют видео. Теперь это мощная база данных, к которой вы можете получить доступ в любое время, просто подумав. Как говорит МакКонахи: «Когда вы пытаетесь вспомнить, что вы ели на завтрак, вы можете задать вопрос: «Что я ел на завтрак?»». Это даст вам буквальную фотографическую память.
И очки, конечно же, будут оснащены каким-то продвинутым чат-ботом, который даст вам прямой доступ ко всем человеческим знаниям и позволит добиваться результатов, просто думая. Допустим, у вас есть проблема, которую необходимо решить. Вы можете быстро и тихо поговорить с этими чат-ботами, которые помогут вам провести мозговой штурм, выработать стратегию, наметить сценарии, оспорить предположения, проверить свои расчеты, подумать, что сказал бы Сократ о вашем затруднительном положении, затем Иммануил Кант, затем Дэвид Юм, а затем Жан-Поль. Сартр.
Теперь ты очень умный. Вы проглотили «безлимитную» таблетку интеллекта. Теперь вы являетесь членом Природы 2.0 и, возможно, наконец-то достигли неизбежного ОИИ. Теперь это честный бой.
Итак, сможет ли блокчейн сыграть какую-то роль в мире сверхлюдей с искусственным интеллектом?
У МакКонахи было похожее видение в 2018 году: идея (как он тогда сказал) «объединить и расширить искусственный интеллект [плюс] блокчейн в симбиозе биологии и машин для будущего изобилия». Суть этой идеи: так же, как сегодняшние данные Web2 принадлежат гигантским компаниям, будущее интеллектуальных данных может быть у нас под рукой. «Как сделать так, чтобы ваши мысли не заметил Google?» — спросил МакКонахи. Какое-то решение на основе блокчейна предоставит вам конфиденциальность, свободу действий и право собственности на ваши собственные мысли, которые теперь подключены к этой более крупной сети.
«По сути, вы должны владеть [своим разумом] так же, как вы владеете биткойнами», — сказал МакКонахи. И это оставляет меня с одной последней мыслью: «Не твои ключи, не твой разум».
Эта история первоначально появилась на Coindesk.