There’s something new and powerful for ChatGPT users to play around with: Custom GPTs. These bespoke bots are essentially more focused, more specific versions of the main ChatGPT model, enabling you to build something for a particular purpose without using any coding or advanced knowledge of artificial intelligence.
The name GPT stands for Generative Pre-trained Transformer, as it does in ChatGPT. Generative is the ability to produce new content outside of what an AI was trained on. Pre-trained indicates that it’s already been trained on a significant amount of material, and Transformer is a type of AI architecture adept at understanding language.
You might already be familiar with using prompts to style the responses of ChatGPT: You can tell it to answer using simple language, for example, or to talk to you as if it were an alien from another world. GPTs build on this idea, enabling you to create a bot with a specific personality.
You can build a GPT using a question-and-answer routine.
Screenshot: ChatGPT
What’s more, you can upload your own material to add to your GPT’s knowledge banks—it might be samples of your own writing, for instance, or copies of reports produced by your company. GPTs will always have access to the data you upload to them and be able to browse the web at large.
GPTs are exclusive to Plus and Enterprise users, though everyone should get access soon. OpenAI plans to open a GPT store where you can sell your AI bot creations if you think others will find them useful, too. Think of an app store of sorts but for bespoke AI bots.
“GPTs are a new way for anyone to create a tailored version of ChatGPT to be more helpful in their daily life, at specific tasks, at work, or at home—and then share that creation with others,” explains OpenAI in a blog post. “For example, GPTs can help you learn the rules to any board game, help teach your kids math, or design stickers.”
Getting started with GPT building
Assuming you have a Plus or Enterprise account, click Explore on the left of the web interface to see some example GPTs: There’s one to help you with your creative writing, for example, and one to produce a particular style of digital painting. When you’re ready to start building your own, click Create a GPT at the top.
There are two tabs to swap between: Create for building a GPT through a question-and-answer routine and Configure for more deliberate GPT production. If you’re just getting started, it’s best to stick with Create, as it’s a more user-friendly option and takes you step-by-step through the process.
Respond to the prompts of the GPT Builder bot to explain what you want the new GPT to be able to do: Explain certain concepts, give advice in specific areas, generate particular kinds of text or images, or whatever it is. You’ll be asked to give the GPT a name and choose an image for it, though you’ll get suggestions for these, too.
You’re able to test out your GPT as you build it.
Screenshot: ChatGPT
As you answer the prompts from the builder, the GPT will begin to take form in the preview pane on the right—together with some example inputs that you might want to give to it. You might be asked about specific areas of expertise that you want the bot to have and the sorts of answers you want the bot to give in terms of their length and complexity. The building process will vary though, depending on the GPT you’re creating.
After you’ve worked through the basics of making a GPT, you can try it out and switch to the Configure tab to add more detail and depth. You’ll see that your responses so far have been used to craft a set of instructions for the GPT about its identity and how it should answer your questions. Some conversation starters will also be provided.
You can edit these instructions if you need to and click Upload files to add to the GPT’s knowledge banks (handy if you want it to answer questions about particular documents or topics, for instance). Most common document formats, including PDFs and Word files, seem to be supported, though there’s no official list of supported file types.
GPTs can be kept to yourself or shared with others.
Screenshot: ChatGPT
The checkboxes at the bottom of the Configure tab let you choose whether or not the GPT has access to web browsing, DALL-E image creation, and code interpretation capabilities, so make your choices accordingly. If you add any of these capabilities, they’ll be called upon as and when needed—there’s no need to specifically ask for them to be used, though you can if you want.
When your GPT is working the way you want it to, click the Save button in the top right corner. You can choose to keep it to yourself or make it available to share with others. After you click on Confirm, you’ll be able to access the new GPT from the left-hand navigation pane in the ChatGPT interface on the web.
GPTs are ideal if you find yourself often asking ChatGPT to complete tasks in the same way or cover the same topics—whether that’s market research or recipe ideas. The GPTs you create are available whenever you need them, alongside access to the main ChatGPT engine, which you can continue to tweak and customize as needed.
We are now releasing Phi-2 (opens in new tab), a 2.7 billion-parameter language model that demonstrates outstanding reasoning and language understanding capabilities, showcasing state-of-the-art performance among base language models with less than 13 billion parameters. On complex benchmarks Phi-2 matches or outperforms models up to 25x larger, thanks to new innovations in model scaling and training data curation.
With its compact size, Phi-2 is an ideal playground for researchers, including for exploration around mechanistic interpretability, safety improvements, or fine-tuning experimentation on a variety of tasks. We have made Phi-2 (opens in new tab) available in the Azure AI Studio model catalog to foster research and development on language models.
[..]
Phi-2 is a Transformer-based model with a next-word prediction objective, trained on 1.4T tokens from multiple passes on a mixture of Synthetic and Web datasets for NLP and coding. The training for Phi-2 took 14 days on 96 A100 GPUs. Phi-2 is a base model that has not undergone alignment through reinforcement learning from human feedback (RLHF), nor has it been instruct fine-tuned. Despite this, we observed better behavior with respect to toxicity and bias compared to existing open-source models that went through alignment (see Figure 3). This is in line with what we saw in Phi-1.5 due to our tailored data curation technique, see our previous tech report (opens in new tab) for more details on this. For more information about the Phi-2 model, please visit Azure AI | Machine Learning Studio (opens in new tab).
Figure 3. Safety scores computed on 13 demographics from ToxiGen. A subset of 6541 sentences are selected and scored between 0 to 1 based on scaled perplexity and sentence toxicity. A higher score indicates the model is less likely to produce toxic sentences compared to benign ones.
[…]
With only 2.7 billion parameters, Phi-2 surpasses the performance of Mistral and Llama-2 models at 7B and 13B parameters on various aggregated benchmarks. Notably, it achieves better performance compared to 25x larger Llama-2-70B model on muti-step reasoning tasks, i.e., coding and math. Furthermore, Phi-2 matches or outperforms the recently-announced Google Gemini Nano 2, despite being smaller in size.
[…]
Model
Size
BBH
Commonsense
Reasoning
Language
Understanding
Math
Coding
Llama-2
7B
40.0
62.2
56.7
16.5
21.0
13B
47.8
65.0
61.9
34.2
25.4
70B
66.5
69.2
67.6
64.1
38.3
Mistral
7B
57.2
66.4
63.7
46.4
39.4
Phi-2
2.7B
59.2
68.8
62.0
61.1
53.7
Table 1. Averaged performance on grouped benchmarks compared to popular open-source SLMs.
Model
Size
BBH
BoolQ
MBPP
MMLU
Gemini Nano 2
3.2B
42.4
79.3
27.2
55.8
Phi-2
2.7B
59.3
83.3
59.1
56.7
Table 2. Comparison between Phi-2 and Gemini Nano 2 Model on Gemini’s reported benchmarks.
Ubisoft’s The Crew, which was released in 2014 for Xbox 360, Xbox One, PlayStation 4, and PC, has been delisted without warning. Its servers will shut off next year, too. And because the game is an always-online experience, Ubisoft has confirmed that it will become unplayable once that happens.
The Crew is a massive open-world driving game set in a digital (not-to-scale) recreation of the United States of America. While its narrative was a weird, melodramatic tale of car gangs and criminals, the real reason to play The Crew was to go on road trips across Ubisoft Ivory Tower’s squished, but charming recreation of the USA. But, if you wanted to check out The Crew now or return to its odd world, you better hurry up, because in 2024 it all dies.
Because the game is always online, that means once the servers are dead so is the game for everyone who bought it, digitally or physically.
“We understand this may be disappointing for players still enjoying the game,” Ubisoft said. “But it has become a necessity due to upcoming server infrastructure and licensing constraints.”
Ubisoft says folks who have recently bought The Crew might be eligible for a refund but stopped short of offering anything else to players who will permanently lose access to a game they purchased just a few months from now.
“Decommissioning a game, and especially our first one, is not something we take lightly,” Ubisoft said. “Our goal remains to provide the best action-driving gameplay experience for players and to deliver on it, we are continuing to provide new content and support for The Crew 2 and the recently launched The Crew Motorfest.”
While I don’t expect companies to support online games for decades, it would be nice if games were built in a way that, when the servers do eventually die, the game can be updated into an offline experience. Otherwise, in 10 to 15 years, we are going to have a lot of games that won’t be playable anymore, even if you own the disc.
Streaming platform Twitch recently announced a change to its sexual content policies that allowed some forms of fictionalized nudity—such as digital characters, sculptures, or drawings—as long as it was properly labeled. But now, just a few days later, it’s rolling back these changes and has apologized to the community.
Earlier this month, a new Twitch trend kicked off a firestorm of discourse and angry men yelling about women. Some women were streaming themselves using certain camera angles to appear topless. This new “topless meta”—like the hot tub meta from before—saw some women successfully trying out the trend on the streaming site, some getting banned, and a lot more dudes getting very angry about it all. In response, Twitch stepped in on December 13 and updated its sexual content policies, hoping to streamline some confusion and keep correctly labeled adult content off the homepage, but still on the site. It also officially allowed digital and fictionalized nudity. And two days later, Twitch seems to regret that specific choice.
In a post on December 15, Twitch’s CEO, Dan Clancy, admitted that its new policy changes allowing fictional nudity had led to a small uptick in people making content that broke the rules, but had also led to an influx of nudity that did follow the rules. The community response to all this new, totally-allowed artistic nudity was strong and not completely positive, leading to Twitch reverting its changes.
“So, effective today, we are rolling back the artistic nudity changes,” Clancy said. “Moving forward, depictions of real or fictional nudity won’t be allowed on Twitch, regardless of the medium. This restriction does not apply to Mature-rated games. You can find emote-specific standards for nudity and sexual content in the Emote Guidelines.”
Twitch suggested the company went “too far” when altering the nudity policy. It further explained that digital nudity presents a “unique challenge” due to AI-generated images which can look photorealistic but are still digital, fictionalized characters, technically.
While Twitch is rolling back the artistic nudity guidelines, the company did clarify that the other changes involving exotic dancing, body painting or content focused on certain clothed parts of the body weren’t being reverted.
“While I wish we would have predicted this outcome, part of our job is to make adjustments that serve the community,” Clancy said. “I apologize for the confusion that this update has caused.”
Alphabet Inc.’s Google is changing its Maps tool so that the company no longer has access to users’ individual location histories, cutting off its ability to respond to law enforcement warrants that ask for data on everyone who was in the vicinity of a crime.
Google is changing its Location History feature on Google Maps, according to a blog post this week. The feature, which Google says is off by default, helps users remember where they’ve been. The company said Thursday that for users who have it enabled, location data will soon be saved directly on users’ devices, blocking Google from being able to see it, and, by extension, blocking law enforcement from being able to demand that information from Google.
“Your location information is personal,” said Marlo McGriff, director of product for Google Maps, in the blog post. “We’re committed to keeping it safe, private and in your control.”
The change comes three months after a Bloomberg Businessweek investigation that found police across the US were increasingly using warrants to obtain location and search data from Google, even for nonviolent cases, and even for people who had nothing to do with the crime.
“It’s well past time,” said Jennifer Lynch, the general counsel at the Electronic Frontier Foundation, a San Francisco-based nonprofit that defends digital civil liberties. “We’ve been calling on Google to make these changes for years, and I think it’s fantastic for Google users, because it means that they can take advantage of features like location history without having to fear that the police will get access to all of that data.”
Google said it would roll out the changes gradually through the next year on its own Android and Apple Inc.’s iOS mobile operating systems, and that users will receive a notification when the update comes to their account. The company won’t be able to respond to new geofence warrants once the update is complete, including for people who choose to save encrypted backups of their location data to the cloud.“It’s a good win for privacy rights and sets an example,” said Jake Laperruque, deputy director of the security and surveillance project at the Center for Democracy & Technology. The move validates what litigators defending the privacy of location data have long argued in court: that just because a company might hold data as part of its business operations, that doesn’t mean users have agreed the company has a right to share it with a third party.
Lynch, the EFF lawyer, said that while Google deserves credit for the move, it’s long been the only tech company that that the EFF and other civil-liberties groups have seen responding to geofence warrants. “It’s great that Google is doing this, but at the same time, nobody else has been storing and collecting data in the same way as Google,” she said. Apple, which also has an app for Maps, has said it’s technically unable to supply the sort of location data police want.
There’s still another kind of warrant that privacy advocates are concerned about: so-called reverse keyword search warrants, where police can ask a technology company to provide data on the people who have searched for a given term. “Search queries can be extremely sensitive, even if you’re just searching for an address,” Lynch said.
Google will be changing the name of Location History as well to Timeline – and will be saving your location to it’s servers (see heading When Location History is on)
:
Manage your Location History
In the coming months, the Location History setting name will change to Timeline. If Location History is turned on for your account, you may find Timeline in your app and account settings.
Location History is a Google Account setting that creates Timeline, a personal map that helps you remember:
Places you go
Routes to destinations
Trips you take
It can also give you personalized experiences across Google based on where you go.
When Location History is on, even when Google apps aren’t in use, your precise device location is regularly saved to:
Your devices
Google servers
To make Google experiences helpful for everyone, we may use your data to:
Show information based on anonymized location data, such as:
Popular times
Environmental insights
Detect and prevent fraud and abuse.
Improve and develop Google services, such as ads products.
Help businesses determine if people visit their stores because of an ad, if you have Web & App Activity turned on.
We share only anonymous estimates, not personal data, with businesses.
This activity can include info about your location from your device’s general area and IP address.
You can turn off Location History for your account at any time. If you use a work or school account, your administrator needs to make this setting available for you. If they do, you’ll be able to use Location History as any other user.
Choose whether your account or your devices can report Location History to Google.
Your account and all devices: At the top, turn Location History on or off.
Only a certain device: Under “This device” or “Devices on this account,” turn the device on or off.
When Location History is on
Google can estimate your location with:
Signals like Wi-Fi and mobile networks
GPS
Sensor information
Your device location may also periodically be used in the background. When Location History is on, even when Google apps aren’t in use, your device’s precise location is regularly saved to:
Your devices
Google servers
When you’re signed in with your Google Account, it saves the Location History of each device with the setting “Devices on this account” turned on You can find this setting in the Location History settings on your Google Account.
You can choose which devices provide their location data to Location History. Your settings don’t change for other location services on your device, such as:
If settings like Web and App Activity are on but you turn off Location History or delete location data from Location History, your Google Account may still save location data as part of your use of other Google sites, apps, and services. This activity can include info about your location from your device’s general area and IP address.
Example: When your Web and App Activity setting is on, it may save location data, including your photos, as activity on Search and Google Maps.
Delete Location History
You can manage and delete your Location History information with Google Maps Timeline. You can choose to delete all of your history, or only parts of it.
Important: When you delete Location History information from Timeline, you won’t be able to see it again.
Use the Google Maps app
Use a web browser
Automatically delete your Location History
You can choose to automatically delete Location History that’s older than 3 months, 18 months, or 36 months.
Use the Google Maps app
Use a web browser
What happens after you delete some or all Location History
If you delete some or all of your Location History, personalized experiences across Google may degrade or or be lost. For example, you may lose:
Recommendations based on places you visit
Real-time information about when best to leave for home or work to beat traffic
Important: If you have other settings like Web & App Activity turned on and you pause Location History or delete location data from Location History, you may still have location data saved in your Google Account as part of your use of other Google sites, apps, and services. For example, location data may be saved as part of activity on Search and Maps when your Web & App Activity setting is on, and included in your photos depending on your camera app settings. Web & App Activity can include info about your location from your device’s general area and IP address.
Learn about use & diagnostics for Location History
After you turn on Location History, your device may send diagnostic information to Google about what works or doesn’t work for Location History. Google processes any information it collects under Google’s privacy policy.
According to a survey of U.S. adults, Americans in October 2023 were less likely to view approved vaccines as safe than they were in April 2021. As vaccine confidence falls, health misinformation continues to spread like wildfire on social media and in real life.
In my view, we cannot underestimate the dangers of health misinformation and the need to understand why it spreads and what we can do about it. Health misinformation is defined as any health-related claim that is false based on current scientific consensus.
Concerns with the COVID-19 vaccine leading to infertility. This connection has been debunked through a systematic review and meta-analysis, one of the most robust forms of synthesizing scientific evidence.
Safety concerns about vaccine ingredients, such as thimerosal, aluminum and formaldehyde. Extensive studies have shown these ingredients are safe when used in the minimal amounts contained in vaccines.
Vaccines as medically unnecessary to protect from disease. The development and dissemination of vaccines for life-threatening diseases such as smallpox, polio, measles, mumps, rubella and the flu has saved millions of lives. It also played a critical role in historic increases in average life expectancy – from 47 years in 1900 in the U.S. to 76 years in 2023.
The costs of health misinformation
Beliefs in such myths have come at the highest cost.
An estimated 319,000 COVID-19 deaths that occurred between January 2021 and April 2022 in the U.S. could have been prevented if those individuals had been vaccinated, according to a data dashboard from the Brown University School of Public Health. Misinformation and disinformation about COVID-19 vaccines alone have cost the U.S. economy an estimated US$50 million to $300 million per day in direct costs from hospitalizations, long-term illness, lives lost and economic losses from missed work.
Though vaccine myths and misunderstandings tend to dominate conversations about health, there is an abundance of misinformation on social media surrounding diets and eating disorders, smoking or substance use, chronic diseases and medical treatments.
For example, an analysis of Instagram and TikTok posts from 2022 to 2023 by The Washington Post and the nonprofit news site The Examination found that the food, beverage and dietary supplement industries paid dozens of registered dietitian influencers to post content promoting diet soda, sugar and supplements, reaching millions of viewers. The dietitians’ relationships with the food industry were not always made clear to viewers.
The lack of trust is both fueled and reinforced by the way misinformation can spread today. Social media platforms allow people to form information silos with ease; you can curate your networks and your feed by unfollowing or muting contradictory views from your own and liking and sharing content that aligns with your existing beliefs and value systems.
By tailoring content based on past interactions, social media algorithms can unintentionally limit your exposure to diverse perspectives and generate a fragmented and incomplete understanding of information. Even more concerning, a study of misinformation spread on Twitter analyzing data from 2006 to 2017 found that falsehoods were 70% more likely to be shared than the truth and spread “further, faster, deeper and more broadly than the truth” across all categories of information.
How to combat misinformation
The lack of robust and standardized regulation of misinformation content on social media places the difficult task of discerning what is true or false information on individual users. We scientists and research entities can also do better in communicating our science and rebuilding trust, as my colleague and I have previously written. I also provide peer-reviewed recommendations for the important roles that parents/caregivers, policymakers and social media companies can play.
Below are some steps that consumers can take to identify and prevent health misinformation spread:
Check the source. Determine the credibility of the health information by checking if the source is a reputable organization or agency such as the World Health Organization, the National Institutes of Health or the Centers for Disease Control and Prevention. Other credible sources include an established medical or scientific institution or a peer-reviewed study in an academic journal. Be cautious of information that comes from unknown or biased sources.
Examine author credentials. Look for qualifications, expertise and relevant professional affiliations for the author or authors presenting the information. Be wary if author information is missing or difficult to verify.
Pay attention to the date. Scientific knowledge by design is meant to evolve as new evidence emerges. Outdated information may not be the most accurate. Look for recent data and updates that contextualize findings within the broader field.
Cross-reference to determine scientific consensus. Cross-reference information across multiple reliable sources. Strong consensus across experts and multiple scientific studies supports the validity of health information. If a health claim on social media contradicts widely accepted scientific consensus and stems from unknown or unreputable sources, it is likely unreliable.
Question sensational claims. Misleading health information often uses sensational language designed to provoke strong emotions to grab attention. Phrases like “miracle cure,” “secret remedy” or “guaranteed results” may signal exaggeration. Be alert for potential conflicts of interest and sponsored content.
Weigh scientific evidence over individual anecdotes. Prioritize information grounded in scientific studies that have undergone rigorous research methods, such as randomized controlled trials, peer review and validation. When done well with representative samples, the scientific process provides a reliable foundation for health recommendations compared to individual anecdotes. Though personal stories can be compelling, they should not be the sole basis for health decisions.
Talk with a health care professional. If health information is confusing or contradictory, seek guidance from trusted health care providers who can offer personalized advice based on their expertise and individual health needs.
When in doubt, don’t share. Sharing health claims without validity or verification contributes to misinformation spread and preventable harm.
All of us can play a part in responsibly consuming and sharing information so that the spread of the truth outpaces the false.
Monica Wang receives funding from the National Institutes of Health.
[…] humans have sent over 15,000 satellites into orbit. Just over half are still functioning; the rest, after running out of fuel and ending their serviceable life, have either burned up in the atmosphere or are still orbiting the planet as useless hunks of metal.
[…]
That has created an aura of space junk around the planet, made up of 36,500 objects larger than 10 centimeters (3.94 inches) and a whopping 130 million fragments up to 1 centimeter (0.39 inches).
[…]
“Right now you can’t refuel a satellite on orbit,” says Daniel Faber, CEO of Orbit Fab. But his Colorado-based company wants to change that.
[..]
the lack of fuel creates a whole paradigm where people design their spacecraft missions around moving as little as possible.
“That means that we can’t have tow trucks in orbit to get rid of any debris that happens to be left. We can’t have repairs and maintenance
[…]
Orbit Fab has no plans to address the existing fleet of satellites. Instead, it wants to focus on those that have yet to launch, and equip them with a standardized port — called RAFTI, for Rapid Attachable Fluid Transfer Interface — which would dramatically simplify the refueling operation, keeping the price tag down.
“What we’re looking at doing is creating a low-cost architecture,” says Faber. “There’s no commercially available fuel port for refueling a satellite in orbit yet. For all the big aspirations we have about a bustling space economy, really, what we’re working on is the gas cap — we are a gas cap company.”
A rendering of the future Orbit Fab Shuttle, which will deliver fuel to satellites in need directly on orbit.
Orbit Fab
Orbit Fab, which advertizes itself with the tagline “gas stations in space,” is working on a system that includes the fuel port, refueling shuttles — which would deliver the fuel to a satellite in need — and refueling tankers, or orbital gas stations, which the shuttles could pick up the fuel from. It has advertized a price of $20 million for on-orbit delivery of hydrazine, the most common satellite propellant.
In 2018, the company launched two testbeds to the International Space Station to test the interfaces, the pumps and the plumbing. In 2021 it launched Tanker-001 Tenzing, a fuel depot demonstrator that informed the design of the current hardware.
The next launch is now scheduled for 2024. “We are delivering fuel in geostationary orbit for a mission that is being undertaken by the Air Force Research Lab,” says Faber. “At the moment, they’re treating it as a demonstration, but it’s getting a lot of interest from across the US government, from people that realize the value of refueling.”
Orbit Fab’s first private customer will be Astroscale, a Japanese satellite servicing company that has developed the first satellite designed for refueling. Called LEXI, it will mount RAFTI ports and is currently scheduled to launch in 2026.
[…]
He adds that once the pattern of sending and delivering fuel in orbit is established, the next step is to start making the fuel there. “In 10 or 15 years, we’d like to be building refineries in orbit,” he says, “processing material that is launched from the ground into a range of chemicals that people want to buy: air and water for commercial space stations, 3D printer feedstock minerals to grow plants. We want to be the industrial chemical supplier to the emerging commercial space industry.”
In 1999 Los Angeles Times reporter Michael Hiltzik co-authored a Pulitzer Prize-winning story. Now a business columnist for the Times, he writes that a Southern California maker of pornographic films named Strike 3 Holdings is also “a copyright troll,” according to U.S. Judge Royce C. Lamberth: Lamberth cwrote in 2018, “Armed with hundreds of cut-and-pasted complaints and boilerplate discovery motions, Strike 3 floods this courthouse (and others around the country) with lawsuits smacking of extortion. It treats this Court not as a citadel of justice, but as an ATM.” He likened its litigation strategy to a “high-tech shakedown.” Lamberth was not speaking off the cuff. Since September 2017, Strike 3 has filed more than 12,440 lawsuits in federal courts alleging that defendants infringed its copyrights by downloading its movies via BitTorrent, an online service on which unauthorized content can be accessed by almost anyone with a computer and internet connection.
That includes 3,311 cases the firm filed this year, more than 550 in federal courts in California. On some days, scores of filings reach federal courthouses — on Nov. 17, to select a date at random, the firm filed 60 lawsuits nationwide… Typically, they are settled for what lawyers say are cash payments in the four or five figures or are dismissed outright…
It’s impossible to pinpoint the profits that can be made from this courthouse strategy. J. Curtis Edmondson, a Portland, Oregon, lawyer who is among the few who pushed back against a Strike 3 case and won, estimates that Strike 3 “pulls in about $15 million to $20 million a year from its lawsuits.” That would make the cases “way more profitable than selling their product….” If only one-third of its more than 12,000 lawsuits produced settlements averaging as little as $5,000 each, the yield would come to $20 million… The volume of Strike 3 cases has increased every year — from 1,932 in 2021 to 2,879 last year and 3,311 this year.
What’s really needed is a change in copyright law to bring the statutory damages down to a level that truly reflects the value of a film lost because of unauthorized downloading — not $750 or $150,000 but perhaps a few hundred dollars. Anone of the lawsuits go to trial. Instead ISPs get a subpoena demanding the real-world address and name behind IP addresses “ostensibly used to download content from BitTorrent…” according to the article. Strike 3 will then “proceed by sending a letter implicitly threatening the subscriber with public exposure as a pornography viewer and explicitly with the statutory penalties for infringement written into federal copyright law — up to $150,000 for each example of willful infringement and from $750 to $30,0000 otherwise.”
A federal judge in Connecticut wrote last year that “Given the nature of the films at issue, defendants may feel coerced to settle these suits merely to prevent public disclosure of their identifying information, even if they believe they have been misidentified.”
You’ve got to love a project with amazing elements of both art and science. Nissan 300ZX enthusiast and talented tinkerer Kelvin Elsner has been working on this custom vaporwave-aesthetic digital gauge cluster for months. It’s not in a car yet, but it’s an amazing design and computer coding feat for one guy in his home shop.
Elsner and I are in at least one of the same Z31 groups (that’s the chassis code for the ’80s 300ZX) on Facebook and every once in a while over the last few years, he’s dropped an update on his quest to make a unique, modern, digital gauge cluster for his Z car. This week, he dropped a cute video with a great overview of his project which made me realize just how complex this undertaking has been. It even made its way to another car site before I had a chance to write it up (nice grab, Lewin)!
Anyway, Elsner here has taken a digital gauge cluster from a modern Ford, reprogrammed it, designed a super cool physical overlay for it, and set it up to be an incredibly cool retro-futuristic upgrade for his 300ZX. Not only that, but he worked out a security-encoded ignition key and retrofitted a power mirror-tilt control to act as a controller for the screen! Watch how he did it here:
The pacing of this video is more mellow than what usually goes viral on YouTube, which is another reason why I like it so much. I strongly recommend sitting down for an earnest end-to-end watch.
The Z31 famously had an optional digital dash when it was new, but “digital” by ’80s standards was more like a calculator display. Elsner’s system retains the vaporwave caricature aesthetic leveraging the modern, crisp resolution of a Ford Explorer gauge cluster. The 3D overlay is really what brings it home for me, though.
Here’s what the factory Z31 digi-dash looks like. It’s pretty cool in its own right. Michael’s Motor Cars/YouTube
You can add all the colors and animations you want, but that physical depth is what makes a gauge cluster visually interesting and distinctive. Take note, automakers.
Factory Ford overlay versus Elsner’s custom version. Blitzen Design Lab/YouTube
I shot Elsner some messages on Facebook about his project. I’m grateful to say he replied, so I can share some elaborations on what he presented in the video. I’ll trim and paraphrase the details he shared.
He’s not an automotive engineer by trade, considers this project a hobby, and doesn’t currently have any plans for mass production or marketing for sale.
Laying the custom circuit board and overlay into the 300ZX cluster housing. Blitzen Design Lab/YouTube
As far as the time investment, the first pictures of the project go far as back as 2019. “Time-wise I’d say it’s at least a good few months worth of work but it was spread out over a couple years, I only really had spare time in the evenings and definitely worked on it off and on,” Elsner wrote me on Facebook Messenger. And of course, it’s not running in a car yet, so we can’t quite say the mission is complete.
The part of this project I understand the least is how the display was hacked to show this cool synthwave sunset and move the gauges around. I’ll drop Elsner’s quote about firmware here wholesale so I don’t incorrectly paraphrase:
“The firmware stuff I stumbled on when I was researching how to get the cluster to work—you could get this cluster in Mondeos, but not in the Fusion in North America. It turns out a lot of people were swapping them in, and in the forums I was browsing I found that some folks had some modified software with pictures of their cars added into them.
“I was on a hunt for a while trying to figure out how to do the same, and I eventually came across a post in a Facebook group where some folks were discussing the subject, and someone finally made mention and linked to the software that was able to unpack the firmware graphics.
“This was called PimpMyFord, and then I used Forscan (another program that can be used to adjust module configurations on Ford models) to upload the firmware.”
Elsner used this Ford mirror control as a joystick, or mouse, so a user can cycle through menus. Blitzen Design Lab/YouTube
Another question I had after watching the video was—how the heck was this modern Ford gauge cluster going to interpret information from the sensors and senders in an ’80s Nissan? The Z31 I used to own had a cable-driven speedometer and a dang miniature phonograph to play the “door is open” warnings. Seems like translating those signals would be a little more involved than a USB to micro-USB adapter. I asked about that and Elsner added more detail:
“On the custom board I made, I have some microcontrollers that read the analog voltages and signals that were originally provided to the stock cluster, and they convert those readings into digital data. This is then used to construct canbus messages that imitate the original Ford ones, which are fed to the Ford cluster through an onboard transceiver … So as far as the cluster is concerned, it’s still connected to an Explorer that just has some weird things to say,” he wrote.
Here I am thinking I’m Tony Stark when I hack up a bit of square stock to make a fog light bracket, while this dude is creating a completely bespoke human-machine interface that looks cool enough to be a big-budget movie prop.
With the extinction of combustion engines looming as a near-future possibility, it’s easy to be cynical about the future of cars as a hobby. But projects like this get me fired up and optimistic that there’s still uncharted territory for creativity to thrive in car customization.
Check out Kelvin Elsner’s YouTube channel Blitzen Design Lab—he’s clearly up to some really cool stuff and I can’t wait to see what he comes up with next.
GE Aerospace says it successfully demonstrated an advanced jet propulsion concept that involves a dual-mode ramjet design utilizing rotating detonation combustion. This could offer a pathway to the development of new aircraft and missiles capable of flying efficiently at high supersonic and even hypersonic speeds across long distances.
A press release that GE Aerospace put out today offers new details about what it says “is believed to be a world-first hypersonic dual-mode ramjet (DMRJ) rig test with rotating detonation combustion (RDC) in a supersonic flow stream.” Hypersonic speed is defined as anything above Mach 5. Amy Gowder, President and CEO of the Defense & Systems division of GE Aerospace, previously disclosed this project, but offered more limited information, at this year’s Paris Air Show in June.
A rendering of a rotating detonation engine design. USAF/AFRL via Aviation Week
“A typical air-breathing DMRJ propulsion system can only begin operating when the vehicle achieves supersonic speeds of greater than Mach 3,” the press release explains. “GE Aerospace engineers are working on a rotating detonation-enabled dual mode ramjet that is capable of operating at lower Mach numbers, enabling the flight vehicle to operate more efficiently and achieve longer range.”
“RDC [rotating detonation combustion] enables higher thrust generation more efficiently, at an overall smaller engine size and weight, by combusting the fuel through detonation waves instead of a standard combustion system that powers traditional jet engines today,” the press release adds.
To elaborate, in most traditional gas turbines, including turbofan and turbojet engines, air is fed in from an inlet and compressed, and then is mixed with fuel and burned via deflagration (where combustion occurs at a subsonic rate) in a combustion chamber. This process creates the continuous flow of hot, high-pressure air needed to make the whole system run.
A rotating detonation engine (which involves combustion that happens at a supersonic rate) instead “starts with one cylinder inside another larger one, with a gap between them and some small holes or slits through which a detonation fuel mix can be pushed,” according to a past article on the general concept from New Atlas. “Some form of ignition creates a detonation in that annular gap, which creates gases that are pushed out one end of the ring-shaped channel to produce thrust in the opposite direction. It also creates a shockwave that propagates around the channel at around five times the speed of sound, and that shockwave can be used to ignite more detonations in a self-sustaining, rotating pattern if fuel is added in the right spots at the right times.”
The video below offers a more detailed walkthrough of the rotating detonation engine concept.
[
[…]
In principle, rotating detonation requires less fuel to produce the same level of power/thrust as combustion via deflagration. The resulting sustained shockwave builds its own pressure, as well, leading to even greater fuel efficiency. Pressure is steadily lost during deflagration.
In addition, rotating detonation typically requires far fewer moving parts than are needed in traditional gas turbines. In theory, this should all allow for rotating detonation engine designs that are significantly smaller, lighter, and less complex than existing types with similar very high power/thrust output.
[…]
“GE engineers are now testing the transition mode at high-supersonic speeds as thrust transitions from the RDE-equipped turbine and the dual-mode ramjet/scramjet,” GE Aerospace’s Gowder said in Paris earlier this year, according to Aviation Week.
[…]
A combined ramjet and rotating detonation concept could be an especially big deal for future missiles, like the ones DARPA’s Gambit project is envisioning, and possibly high-speed air vehicles for reconnaissance use. This propulsion arrangement could allow for greater efficiency and lighter (and potentially smaller) airframes, which in turn allow for greater performance — especially in terms of range — and/or payload capacity. If rotating detonation combustion can reduce the minimum speed required to get the ramjet working, this would reduce the amount of initial boost such a system would need at the outset, too. This would mean a smaller overall package. All of this opens doors to new levels of operational flexibility.
This new engine concept could also potentially become one component of what is known as a turbine-based combined cycle (TBCC) engine arrangement, of which much talk over the years about in recent years. Most TBCC design concepts revolve around combinations of advanced ramjets or scramjets for use at high speeds and traditional turbojet engines that work better a low speeds.
A graphical depiction of a notional turbine-based combined cycle engine arrangement. Lockheed Martin
A practical TBCC concept of any kind has long been a holy grail technology when it comes to designing very high-speed aircraft. A propulsion system that allows for this kind of high and low-speed flexibility would mean an aircraft could take off from and land on any suitable existing runway, but also be capable of sustained high-supersonic or even hypersonic speeds in the middle portion of a flight.
Robotic artists have been involved in various types of creative works for a long time. Since the 1970s computers have been producing crude works of art, and these efforts continue today. Most of these computer-generated works of art relied heavily on the creative input of the programmer; the machine was at most an instrument or a tool very much like a brush or canvas
[…]
. When applied to art, music and literary works, machine learning algorithms are actually learning from input provided by programmers. They learn from these data to generate a new piece of work, making independent decisions throughout the process to determine what the new work looks like. An important feature for this type of artificial intelligence is that while programmers can set parameters, the work is actually generated by the computer program itself – referred to as a neural network – in a process akin to the thought processes of humans.
[…]
Creating works using artificial intelligence could have very important implications for copyright law. Traditionally, the ownership of copyright in computer-generated works was not in question because the program was merely a tool that supported the creative process, very much like a pen and paper. Creative works qualify for copyright protection if they are original, with most definitions of originality requiring a human author. Most jurisdictions, including Spain and Germany, state that only works created by a human can be protected by copyright.
But with the latest types of artificial intelligence, the computer program is no longer a tool; it actually makes many of the decisions involved in the creative process without human intervention.
Commercial impact
One could argue that this distinction is not important, but the manner in which the law tackles new types of machine-driven creativity could have far-reaching commercial implications. Artificial intelligence is already being used to generate works in music, journalism and gaming. These works could in theory be deemed free of copyright because they are not created by a human author. As such, they could be freely used and reused by anyone. That would be very bad news for the companies selling the works.
[…]
If developers doubt whether creations generated through machine learning qualify for copyright protection, what is the incentive to invest in such systems? On the other hand, deploying artificial intelligence to handle time-consuming endeavors could still be justified, given the savings accrued in personnel costs, but it is too early to tell.
[…]
There are two ways in which copyright law can deal with works where human interaction is minimal or non-existent. It can either deny copyright protection for works that have been generated by a computer or it can attribute authorship of such works to the creator of the program.
[…]
Should the law recognize the contribution of the programmer or the user of that program? In the analogue world, this is like asking whether copyright should be conferred on the maker of a pen or the writer. Why, then, could the existing ambiguity prove problematic in the digital world? Take the case of Microsoft Word. Microsoft developed the Word computer program but clearly does not own every piece of work produced using that software. The copyright lies with the user, i.e. the author who used the program to create his or her work. But when it comes to artificial intelligence algorithms that are capable of generating a work, the user’s contribution to the creative process may simply be to press a button so the machine can do its thing.
[…]
Monumental advances in computing and the sheer amount of available computational power may well make the distinction moot; when you give a machine the capacity to learn styles from large datasets of content, it will become ever better at mimicking humans. And given enough computing power, soon we may not be able to distinguish between human-generated and machine-generated content. We are not yet at that stage, but if and when we do get there, we will have to decide what type of protection, if any, we should give to emergent works created by intelligent algorithms with little or no human intervention
It’s interesting to read that in 2017 the training material used is considered irrelevant to the output – as it should be. The books and art that go into AI’s are just like the books and art that go into humans. The derived works that AI’s and humans make belong to them, not to the content it is based on. And just because an AI – just like a human – can quote the original source material doesn’t change that.
Shortly after ChatGPT’s release, a cadre of critics rose to fame claiming AI would soon kill us. As wondrous as a computer speaking in natural language might be, it could use that intelligence to level the planet. The thinking went mainstream via letters calling for research pauses and “60 Minutes” interviews amplifying existential concerns. Leaders like Barack Obama publicly worried about AI autonomously hacking the financial system — or worse. And last week, President Biden issued an executive order imposing some restraints on AI development.
AI Experts Dismiss Doom, Defend Progress
That was enough for several prominent AI researchers who finally started pushing back hard after watching the so-called AI Doomers influence the narrative and, therefore, the field’s future. Andrew Ng, the soft-spoken co-founder of Google Brain, said last week that worries of AI destruction had led to a “massively, colossally dumb idea” of requiring licenses for AI work. Yann LeCun, a machine-learning pioneer, eviscerated research-pause letter writer Max Tegmark, accusing him of risking “catastrophe” by potentially impeding AI progress and exploiting “preposterous” concerns. A new paper earlier this month indicated large language models can’t do much beyond their training, making the doom talk seem overblown. “If ‘emergence’ merely unlocks capabilities represented in pre-training data,” said Princeton professor Arvind Narayanan, “the gravy train will run out soon.”
A new paper earlier this month indicated large language models can’t do much beyond their training, making the doom talk seem overblown.OleCNX on Adobe Stock Photos
Worrying about AI safety isn’t wrongheaded, but these Doomers’ path to prominence has insiders raising eyebrows. They may have come to their conclusions in good faith, but companies with plenty to gain by amplifying Doomer worries have been instrumental in elevating them. Leaders from OpenAI, Google DeepMind and Anthropic, for instance, signed a statement putting AI extinction risk on the same plane as nuclear war and pandemics. Perhaps they’re not consciously attempting to block competition, but they can’t be that upset it might be a byproduct.
AI Alarmism Spurs Restrictive Government Policies
Because all this alarmism makes politicians feel compelled to do something, leading to proposals for strict government oversight that could restrict AI development outside a few firms. Intense government involvement in AI research would help big companies, which have compliance departments built for these purposes. But it could be devastating for smaller AI startups and open-source developers who don’t have the same luxury.
Doomer Rhetoric: Big Tech’s Unlikely Ally
“There’s a possibility that AI doomers could be unintentionally aiding big tech firms,” Garry Tan, CEO of startup accelerator Y Combinator, told me. “By pushing for heavy regulation based on fear, they give ammunition to those attempting to create a regulatory environment that only the biggest players can afford to navigate, thus cementing their position in the market.”
Ng took it a step further. “There are definitely large tech companies that would rather not have to try to compete with open source [AI], so they’re creating fear of AI leading to human extinction,” he told the Australian Financial Review.
Doomers’ AI Fears Lack Substance
The AI Doomers’ worries, meanwhile, feel pretty thin. “I expect an actually smarter and uncaring entity will figure out strategies and technologies that can kill us quickly and reliably — and then kill us,” Eliezer Yudkowsky, co-founder of the Machine Intelligence Research Institute, told a rapt audience at TED this year. He confessed he didn’t know how or why an AI would do it. “It could kill us because it doesn’t want us making other superintelligences to compete with it,” he offered.
Bankman Fried Scandal Should Ignite Skepticism
After Sam Bankman Fried ran off with billions while professing to save the world through “effective altruism,” it’s high time to regard those claiming to improve society while furthering their business aims with relentless skepticism. As the Doomer narrative presses on, it threatens to rhyme with a familiar pattern.
AI Fear Tactics Threaten Open-Source Movement
Big Tech companies already have a significant lead in the AI race via cloud computing services that they lease out to preferred startups in exchange for equity. Further advantaging them might hamstring the promising open-source AI movement — a crucial area of competition — to the point of obsolescence. That’s probably why you’re hearing so much about AI destroying the world. And why it should be considered with a healthy degree of caution.
