Tag:

Learning

CrowPi3 Electronic Learning Platform header image.
Product Reviews

CrowPi3 Electronic Learning Platform review

by admin



Why you can trust TechRadar


We spend hours testing every product or service we review, so you can be sure you’re buying the best. Find out more about how we test.

Elecrow’s latest invention, the CrowPi3, is an all-in-one Raspberry Pi 5 platform that bundles many popular Arduino modules in a book-sized suitcase. A 4.3” LCD sits in the center with a camera on top. Modules usually used to introduce electronics, such as a breadboard or an LED matrix, are also included. They are all connected to the Pi through its 40-pin header connector. The unit has a built-in battery pack that should provide hours of fun.

Elecrow is running a Kickstarter campaign that will end on July 5th. The product’s main selling point is that it can be used as a hardware enabler for AI designs. How much this is true depends on the software ecosystem that ships with it. Elecrow is smart by using the Raspberry Pi as part of its design. This opens thousands of AI and ML libraries for the younger generation targeted by this product.

CrowPi3: Price and Availability

The CrowPi3 is available in nine configurations on the Kickstarter webpage. The Early Birds offer the most value for money and start at $159 for the basic kit, excluding the Pi computer. An 8GB Pi 5 with battery and a gamepad will add an extra $130.

(Image credit: Future)

CrowPi3: Design

The CrowPi3 ships with a transparent plastic cover on the main experiment board, which helps protect the delicate circuits and LCD module while on the go.

Specification

Supported board: Raspberry Pi 5, Pi Pico, Arduino, Micro: bit

Number of extensions: 28 Arduino-compatible modules

Size: 28.5 cm x 18.5 cm x 3.8 cm

Weight: 1.1 kg

Video: 4.3” 800 x 480 IPS LCD, 1 x HDMI

Audio: Stereo speaker, 3.5mm headphone jack

Power supply: 27W 12V power adapter

Two small magnets hold the cover in place while a rubber strap on top acts as a handle, creating a look similar to a suitcase. The unit measures 28.5 cm x 18.5 cm x 3.8 cm, for a weight of 1.1 kg, and feels solid. The two compartments on the bottom are easily accessible by sliding plastic doors. One holds the Raspberry Pi board, while the other can store small parts.

Elecrow went above and beyond to pack over thirty electronic modules in a small volume. Even more impressive is that they are all connected to the Pi expansion connector. If the Pi is too much of a hassle to work with, then two extra sockets accommodate a Raspberry Pi Pico board or an Arduino, while the breadboard space allows adding other modules easily.

The Raspberry Pi 5 provides over twice the processing power of its predecessor. The CrowPi3 utilizes all of the Pi’s connectors and interfaces; Two HDMI connectors connect to the integrated display and an optional external screen. The user can access only three USB 3 ports, the fourth reserved for the two-megapixel camera. A 4.3-inch 800 x 480 IPS screen offers excellent brightness and viewing angle, although the small viewing area makes reading text challenging.

The case integrates two 3300mAh rechargeable batteries, which should keep the Pi and screen working for several hours. Networking is available using the 1 Gb Ethernet port or the Pi’s Wi-Fi module. A pair of speakers provides stereo sound output, which can be bypassed using a headphone plugged into the 3.5mm headphone jack.

(Image credit: Future)

CrowPi3: In Use

The unit turns on by pressing the right-side power button. A long press on the same button turns the unit off. Two red LEDs indicate whether the unit is charging and turned on. The cooling system is always on without temperature control and produces a low but noticeable whining sound from the fan. The built-in screen is more of a gimmick. An external display is recommended for actual development.

The Elecrow ecosystem builds on a customized version of Raspbian OS, specially tailored for interactive learning. The UI ditches the conventional desktop with icons and instead organizes items in what looks like a dashboard with two distinct sections. The left contains AI modules, while the right holds Python-based projects. The user must create an account to start using the CrowPi3.

The most challenging part of setting up the CrowPi3 is getting the OS up and running. Elecrow provides a 40GB image that, although containing a wealth of material, is just too big to be useful. In contrast, the latest Raspbian desktop image with recommended applications is less than 4 GB. Elecrow should provide a lite version of the OS with additional user packages external to the image.

(Image credit: Future)

CrowPi3: Competition

Very few platforms exist that are similar to the CrowPi3. With built-in modules that showcase the Pi’s ability to control peripherals and the various extension connectors spread around the case, the CrowPi3 is a unique product. The curated software library also plays an essential part in the product.

The Joy-Pi advanced product is similar to the CrowPi3 and supports Pi 4. It incorporates modules such as touch sensors and ultrasonic range finders and sports an even smaller 1-inch LCD. However, the price is higher than the CrowPi3, at more than $470. It lacks a battery and key components such as a camera to make an AI platform.

(Image credit: Future)

CrowPi3: Verdict

With their newest all-in-one station, Elecrow promotes an AI-centric product on a large scale through their Kickstarter campaign. The CrowPi3 feels well-built and oozes quality with many electronic modules, such as the LCD and touchscreen. However, the software support for voice and image recognition sets it apart from the competition. We liked the general appearance and portability of the platform, with a cute handle that completes the suitcase appearance. Don’t let the toy appearance fool you; the platform can be used for more serious stuff, such as software development, but with a bigger external screen.

That said, the unit feels overcrowded. Elecrow thinks cramming as many modules into a small space will win over the general public. The fact is that the platform might overwhelm beginners with its complexity. Also, supporting material such as schematics is missing. You are on your own if the unit breaks and you want to do some repairs. Finally, the fan is noisy and will be a distraction.

Buy it as a gift for kids who like building and breaking stuff.

Don’t buy it if you want to get your hands dirty quickly with the Pi5.

Raspberry Pi: Everything you need to know.



Source link

0 comments
0 FacebookTwitterPinterestEmail
This AI Model Never Stops Learning
Gaming Gear

This AI Model Never Stops Learning

by admin


Modern large language models (LLMs) might write beautiful sonnets and elegant code, but they lack even a rudimentary ability to learn from experience.

Researchers at Massachusetts Institute of Technology (MIT) have now devised a way for LLMs to keep improving by tweaking their own parameters in response to useful new information.

The work is a step toward building artificial intelligence models that learn continually—a long-standing goal of the field and something that will be crucial if machines are to ever more faithfully mimic human intelligence. In the meantime, it could give us chatbots and other AI tools that are better able to incorporate new information including a user’s interests and preferences.

The MIT scheme, called Self Adapting Language Models (SEAL), involves having an LLM learn to generate its own synthetic training data and update procedure based on the input it receives.

“The initial idea was to explore if tokens [units of text fed to LLMs and generated by them] could cause a powerful update to a model,” says Jyothish Pari, a PhD student at MIT involved with developing SEAL. Pari says the idea was to see if a model’s output could be used to train it.

Adam Zweiger, an MIT undergraduate researcher involved with building SEAL, adds that although newer models can “reason” their way to better solutions by performing more complex inference, the model itself does not benefit from this reasoning over the long term.

SEAL, by contrast, generates new insights and then folds it into its own weights or parameters. Given a statement about the challenges faced by the Apollo space program, for instance, the model generated new passages that try to describe the implications of the statement. The researchers compared this to the way a human student writes and reviews notes in order to aid their learning.

The system then updated the model using this data and tested how well the new model is able to answer a set of questions. And finally, this provides a reinforcement learning signal that helps guide the model toward updates that improve its overall abilities and which help it carry on learning.

The researchers tested their approach on small and medium-size versions of two open source models, Meta’s Llama and Alibaba’s Qwen. They say that the approach ought to work for much larger frontier models too.

The researchers tested the SEAL approach on text as well as a benchmark called ARC that gauges an AI model’s ability to solve abstract reasoning problems. In both cases they saw that SEAL allowed the models to continue learning well beyond their initial training.

Pulkit Agrawal, a professor at MIT who oversaw the work, says that the SEAL project touches on important themes in AI, including how to get AI to figure out for itself what it should try to learn. He says it could well be used to help make AI models more personalized. “LLMs are powerful but we don’t want their knowledge to stop,” he says.

SEAL is not yet a way for AI to improve indefinitely. For one thing, as Agrawal notes, the LLMs tested suffer from what’s known as “catastrophic forgetting,” a troubling effect seen when ingesting new information causes older knowledge to simply disappear. This may point to a fundamental difference between artificial neural networks and biological ones. Pari and Zweigler also note that SEAL is computationally intensive, and it isn’t yet clear how best to most effectively schedule new periods of learning. One fun idea, Zweigler mentions, is that, like humans, perhaps LLMs could experience periods of “sleep” where new information is consolidated.

Still, for all its limitations, SEAL is an exciting new path for further AI research—and it may well be something that finds its way into future frontier AI models.

What do you think about AI that is able to keep on learning? Send an email to hello@wired.com to let me know.



Source link

0 comments
0 FacebookTwitterPinterestEmail
Dogecoin price
NFT Gaming

When Will The Dogecoin Price Hit $1 In 2025? Machine Learning Algorithm Answers

by admin


Trusted Editorial content, reviewed by leading industry experts and seasoned editors. Ad Disclosure

Despite a rollercoaster few years that have seen the Bitcoin price hit multiple new all-time highs, other high-cap cryptocurrencies like Dogecoin have failed to do the same. In fact, the highest the Dogecoin price has gotten in the last three years is $0.5, which is still a ways away from the $0.74 figure that was hit back in 2021. Calls for $1 for the meme coin have also died down as it continues to decline and struggle along with the rest of the altcoin market.

Can Dogecoin Price Reach $1?

When it comes to the Dogecoin price, expectations continue to falter, especially as the price remains low with resistance mounting at $0.2. To reach the $1 mark, the price would have to rally by 500% from its current level, and with trading volumes and interest so low, the possibility of this happening grows dimmer by the day.

In the same vein, the algorithm at CoinCodex has also followed the bearish school of thought, with a drawn-out timeframe for when it expects the Dogecoin price to hit $1. According to the algorithm, the Dogecoin price will not be hitting the $1 level in the year 2025, or anytime soon after that.

The predictions show that the highest level the Dogecoin price could reach this year is $0.289. As the year moves forward, it expects the price to trend lower, not seeing more than a 54% price increase after the month of September.

Nevertheless, it does expect the price to be higher than where it is now, with an average 25% increase from the current level. It also signals the current level as a buy point with the possibility of making around a 35% increase in the month of June.

Source: CoinCodex

DOGE Won’t Hit $1 Until After 2030

The longer-term predictions for the Dogecoin price show more recovery for the price. But reaching the $1 level is not something that the algorithm believes will happen in the next half a decade. In fact, from 2025 through to 2030, it predicts that the Dogecoin price will not rise to its current all-time high of $0.74.

For those waiting for the Dogecoin price to hit $1, it shows that the price will hit this level sometime after 2030, putting it above $1 in 2040. This could mean that the meme coin would take almost a decade to reach the coveted $1 target.

However, in the short term, the algorithm does see some positive prospects for the Dogecoin price. It predicts that DOGE will rise to $0.236 in the next month, which would be a 17.67% increase from the current level.

DOGE bears push below $0.18 support | Source: DOGEUSDT on TradingView.com

Featured image from Dall.E, chart from TradingView.com

Editorial Process for bitcoinist is centered on delivering thoroughly researched, accurate, and unbiased content. We uphold strict sourcing standards, and each page undergoes diligent review by our team of top technology experts and seasoned editors. This process ensures the integrity, relevance, and value of our content for our readers.



Source link

0 comments
0 FacebookTwitterPinterestEmail
Dawn of the Hunt Ascendancies
Product Reviews

Path of Exile 2 director says GGG is learning from its mistakes while supporting 2 games at once: ‘If you want to lock the date, you can’t lock the features’

by admin



When Grinding Gear Games first announced Path of Exile 2, it made a promise: PoE1 wasn’t going anywhere, and development on both titles would continue in parallel. Ten months after the last Path of Exile league, it’s clear that the studio bit off a bit more than it can chew, but game director Jonathan Rogers was optimistic in a new interview with YouTuber Talkative Tri this week that things are close to being back on track.

One of the main problems, as he sees it, was that GGG veered from a development philosophy it’d had for a long time.

“If you want to lock the features, you can’t lock the date,” he said. “And if you want to lock the date, you can’t lock the features. In the past, we were always locking dates, but then as we had PoE2 in development, we got to the point where we were locking features instead and I think honestly that wasn’t the best.”


You may like

Path of Exile 2’s most recent update was called Dawn of the Hunt, and it’s clear that the focus was squarely on a new class, the Huntress. This painted GGG into a bit of a corner—you can’t really have a hunt dawning without, you know, someone to do the hunting. They couldn’t make announcements about it until they were sure that she was ready, let alone start talking about anything they had cooking with PoE1. As a result, the update felt rushed, and personally I think that was a big contributor to its cool reception.

Moving forward, Rogers said, this uncertainty will be a thing of the past. “We’re doing four months, it’s going to be consistent, it’s always going to be the case. We don’t wanna have this kind of uncertainty around this any more,” he declared, to the resounding cheers of grinders everywhere. With the new 3.26 league coming June 13 for PoE1, this puts us squarely on the release cadence initially promised.

(Image credit: Grinding Gear Games)

A new league or major patch every two months sounds like a dream come true for a PoE sicko like myself, but after over 300 days of radio silence on PoE1’s 3.26 update I think you’d forgive me for being just a touch skeptical. I’m thinking we might see that number stretch a little bit, particularly at first.

Rogers was adamant, though: “With the new development model, like the date is fixed and the content is not, that might mean that the class we’re planning to add might not make it. Now I think it will, but if it had to not make it in order to make it so all the other classes we already have are good, then we will do that tradeoff and that is not the mindset we were in for 0.2.0.”

He added that their primary focus for PoE2’s 0.3 update was making sure that existing classes felt good to play, and if that came at the expense of the new one, so be it. The willingness to make cuts in the name of deadlines tells me that GGG is serious about sticking to the schedule, which is something the community’s been clamoring for since the usual release window for PoE1 came and went half a year ago.

(Image credit: Grinding Gear Games)

We’ll have to see how things shake out with the schedule, but for now at least the content is heading in the right direction. There have been a number of patches to Path of Exile 2 even in this interstitial time between major updates, improving some of the loot issues and other problems that Dawn of the Hunt had. Patch 0.2.1 should be dropping any second now, which has further loot improvements, a whole-ass Expedition revamp, and new chase uniques. It also fixes some major issues—I knew Hidden Grotto was a pain, but did you know that it was 37.5% of all maps on the atlas? And apparently the chaos flowers in Ritual were responsible for fifty percent of player deaths in the endgame? Yeesh.

GGG is hosting a livestream with details for the new PoE1 league on June 5, where we’ll learn more about the patch.



Source link

0 comments
0 FacebookTwitterPinterestEmail
Dogecoin
Crypto Trends

How Will The Dogecoin Price Close Out May? Machine Learning Algorithm Leans Bearish

by admin


Trusted Editorial content, reviewed by leading industry experts and seasoned editors. Ad Disclosure

The Dogecoin price is in focus as the month draws to a close. The meme coin has already enjoyed a 30% gain so far in May, but this machine learning algorithm suggests that DOGE is unlikely to enjoy further gains until the end of the month. 

Machine Learning Algorithm Predicts Bearish Ending For Dogecoin Price

Coincodex’s machine learning algorithm has predicted that the Dogecoin price will suffer more declines in the coming days to close the month. The algorithm predicts that DOGE will drop to $0.220052 by May 28 and then suffer another decline to $0.217269 by May 29. The top meme coin will then drop to $0.214764 and $0.213578 by May 30 and 31, respectively. 

Cryptorank data shows that the Dogecoin price has already gained over 30% in May, having rallied above the psychological $0.2 level earlier this month following Bitcoin’s surge. However, based on the machine learning algorithm’s prediction, DOGE is more likely to trim these gains before the month ends. 

Regardless, the Dogecoin price is on course to record its best monthly performance so far this year. Barring any massive crash, it will also be the first time this year that DOGE has witnessed a positive monthly close with double-digit gains. The meme coin recorded a 4% and 3% gain in January and April, respectively, the two months this year in which it has recorded a positive monthly close. 

Meanwhile, based on historical data, June will likely be a bearish month for the Dogecoin price. DOGE has witnessed negative monthly closes in June in the last eight years. In fact, the meme coin has only recorded two positive monthly closes in June since its creation. However, both were double-digit gains, meaning that Dogecoin could record massive gains if this June ends up being a green month. 

DOGE RSI Indicates Imminent Pump

In an X post, crypto analyst Trader Tardigrade stated that the Dogecoin price’s Relative Strength Index (RSI) suggests a potential upcoming pump for the meme coin. His accompanying chart showed that DOGE could rally to as high as $0.9 on this breakout to the upside. In another post, he affirmed that the meme coin is ready for blast off, which would send its price to a new all-time high (ATH). 

Source: Trader Tardigrade on X
Crypto analyst Ali Martinez highlighted the range between $0.24 and $0.26 as the most important resistance wall for the Dogecoin price. He suggested that a break above this range could spark a parabolic rally for the meme coin, with the current local high at around $0.46 in sight. 

At the time of writing, the Dogecoin price is trading at around $0.22, down almost 2% in the last 24 hours, according to data from CoinMarketCap.

DOGE trading at $0.22 on the 1D chart | Source: DOGEUSDT on Tradingview.com

Featured image from Getty Images, chart from Tradingview.com

Editorial Process for bitcoinist is centered on delivering thoroughly researched, accurate, and unbiased content. We uphold strict sourcing standards, and each page undergoes diligent review by our team of top technology experts and seasoned editors. This process ensures the integrity, relevance, and value of our content for our readers.



Source link

0 comments
0 FacebookTwitterPinterestEmail
Close