Fact check

Earlier this week, on the heels of guilty verdicts in the Steubenville case of two football players who sexually assaulted a girl and broadcast it on social media, teenagers in the town of Torrington, Connecticut, were blowing up Facebook and Twitter with vile, victim-blaming comments about a 13-year-old who says she was raped by two 18-year-old football players. If we hadn’t been aware before, this week has made clear that, in the digital era, the accuser and the accused are not the only parties involved in sexual assault cases. We’re all bystanders on social networks.

Social media has been rightly hailed for bringing the Steubenville crime to light. Texts, tweets, and photos were essential to establishing that this was not a consensual act and not a “he said/she said” story. Yet paradoxically, both the Steubenville and Torrington rape cases have escalated the narrative — one that’s swirled around ever since the days of A/S/L queries in an AOL chatrooms — that the Internet is jeopardizing the privacy and safety of teens. And, in particular, teen girls.

Before it was even available in every home, let alone every purse and pocket in America, adults were worried that the Internet would enable sexual assault and stalking of naïve teens, who would log on and reveal details about themselves to older predators. As we welcomed Web 2.0 and smartphones and broadband into our lives, it became clear that their peers perhaps posed a greater risk, as social media became the place where bullying and rumor-mongering — practices as old as schoolyard itself — became easy to spread at the click of a button. The case of Amanda Todd took these fears to a horrifying extreme: The Canadian teenager was so brutally harassed after a topless photo of her made the rounds in her high school that she eventually committed suicide.

Even though the Steubenville case involved the malicious spread of information online — photos of the passed-out victim were plastered all over social media without her consent — the fact that the case ended in two rape convictions notably revealed that social media is not always negative for young women. “Cellphone videos can be forwarded to authorities, not circulated as jokes,” writes Amanda Hess at Slate. “Text messages can be used to identify rapists, not shame victims. And photos can establish central facts, not publicize humiliation.”

When issuing the Steubenville verdict, Judge Thomas Lipps warned teens to think about “how you record things on the social media so prevalent today.” Lipps, with his finger-wagging tone, is not alone in his fears about the effects of Instagram and Twitter and Facebook. As columnist Kathleen Parker writes this week in the Washington Post , “What hasn’t been addressed is the factor of social media in the events themselves.” She wonders whether our tendency to ‘gram and tweet the tiniest details has detached us from events unfolding in front of our faces, some modern form of the psychological effect that murder victim Kitty Genovese ushered in almost 60 years ago. The implication is that all of the high schoolers who tweeted and retweeted and texted the violence that was perpetrated against this girl in Ohio would have somehow been more likely to intervene in a pre-digital era. Decades of research on the Bystander Effect has shown us that’s not true, which means the Internet is once again acquitted. Bullies, killers, and rapists are the problem, not the medium by which they broadcast their crimes.

In almost every prominent news story about how the dynamics of social media play out offline, young women are set up as victims rather than agents and drivers of technology. The case has been made repeatedly that the digital era puts young women at risk in new ways — they can be stalked with smartphones, slut-shamed on instant messenger, targeted by rapists on social media. But the Internet can also be a source of power and protection — and I’m not just talking about accountability in rape cases. (A new report by the Pew Internet and American Life Project finds that 34 percent of teen girls “mostly go online using their cell phone,” compared with 24 percent of teen boys. “This is notable since boys and girls are equally likely to be smartphone owners,” the study says.) I use an app called CheckOn.Me , which notifies a select group of contacts if I fail to log in to the app after, say, meeting up with some unknown CraigsList seller about buying a chair. More generally, social media allows me a way to casually check up on friends. If I know a friend of mine was on a blind date the night before, I confess that I am slightly relieved when I see a new tweet from her in the morning or that she’s liked something on Facebook. For all the headlines about women being creeped online, it’s easy to forget that, far more often, social media provides new ways to blow the whistle.

The Internet isn’t some lawless netherworld, it’s merely a reflection of social dynamics — and yes, sometimes crimes — that occur in the offline world. Put another way, by Sarah Gram in an essay about teen girls and selfies, “Do we honestly think that by ceasing to take and post selfies, the bodies of young women would cease to be spectacles?” If we didn’t have Instagram pictures and tweets about it, the rape that occurred in Steubenville wouldn’t be any less real. It just would have been less documented. I agree with Judge Lipps that we should all think about how we’re using social media: Rather than just recognizing its potential to hurt and expose, start appreciating how it enables us to help keep each other safe.

Construction is officially underway on the last phase of Auckland’s billion-dollar Eastern Busway .

The first part of the free-flowing lane network has already connected Panmure to Pakuranga, and the next section will connect Botany to the existing network.

The project has had numerous ups and downs as funding shortfalls have delayed the construction moving ahead.

But in December last year the Government announced a $200 million contribution to the project, which will also see a new station built at Botany .

READ MORE: Road cones and bus trials: Fact-checking Auckland mayor Wayne Brown’s big ideas Aucklanders to get no say on mega transport agreement with government

Auckland mayor Wayne Brown said he had pushed to get the “vital project” underway.

The $1.4b project will bring rapid transit to the sprawling eastern suburbs, with the dedicated busway and cycle lanes out to the terminus in Botany.

The new busway will give commuters a high-frequency service every few minutes in peak periods, cutting the travel time from Botany Town Centre to Britomart by 20 minutes.

Around 7,000 people are already using the Panmure to Pakuranga busway daily, according to Auckland Transport.

Brown said Auckland Transport expected 14,000 passengers a day to use the busway by 2028, and 23,000 from 2048.

“It is a great example of a project that provides Aucklanders with faster journeys, both for people catching buses and for people driving,” Brown said.

The project also includes a Reeves Rd flyover to split traffic coming off the Waipuna Bridge from Ti Rakau Drive, creating a direct link to Pakuranga Rd.

A council funding squeeze in 2021 pushed the expected completion date of the busway out by a further two years from its initial end date.

The Integrated Auckland Transport Plan (IATP) was announced in December 2022, and hailed by mayor Wayne Brown as the first time that all projects, and the possible port relocation , were in one conversation.

The draft plan will be ready by the end of May, steered by both the mayor, and the minister for Transport Michael Wood.

Bottom line: Hackers hacked a Tesla Model 3 using a trinity of exploits. Tesla says that the weaknesses only allowed the attackers to operate non-vital functions and that they could not use them to gain access to the engine or other driving features. The security researchers say they think they could but don’t have proof yet. The back and forth is all a moot point, though, since Tesla has patches on the way.

Last week, researchers at Pwn3Own 2023 found three vulnerabilities in a Tesla Model 3 to win $140,000 and the car. The exploits allowed the team from security firm Synacktiv to control some functions of the vehicle remotely, including operating the lights, horn, windshield wipers, infotainment center, and opening the trunk lid.

The first vulnerability was a Bluetooth hack that gave them a foot in the door. The second was an exploit that granted the hackers root privileges to at least one of the Model 3’s systems allowing them to execute arbitrary code. The third weakness compromised the “security gateway,” which handles some commands sent to the car.

As is customary for the Pwn2Own event, host Trend Micro informed Tesla of the zero-day exploits so it would have a chance to plug the holes before publicly announcing the details of the hacks about 90 days from now. Tesla told Synacktiv that although its team could access some rudimentary functions that would, at worst, annoy the owner, it still would not have been able to execute vital functions like turning the engine on or operating the steering wheel.

However, Eloi Benoist-Vanderbeken, one of Synacktiv’s reverse engineers, indicated that Tesla’s assumption might not be valid.

“[Tesla] said we wouldn’t be able to turn the steering wheel, accelerate, or brake. But from our understanding of the car architecture, we are not sure that this is correct, but we don’t have proof of it,” Benoist-Vanderbeken told TechCrunch.

According to the security expert, the team does not have access to a Tesla, even though it won the hacked Model 3 at the event. He did not mention why they don’t have the car but noted that his team looks forward to fact-checking Tesla’s claim.

While Tesla has not commented publically on the vulnerabilities, it indicated that its developers are working on patches that should roll out in an over-the-air update soon. To the car company’s credit, the Synacktiv team said Tesla is “doing a good job” of hardening its systems.

One of the team’s trickier roadblocks was a “mature” system of sandboxes that locks off one component from another. Such compartmentalization prevents attackers from accessing one system by compromising another. Synacktiv Cyber Security Engineer Vincent Dehors compared Tesla’s security to mobile web browsers.

“It’s not at the point of a modern browser running on an iPhone or an Android, but it’s not that far from it,” Dehors said. “Tesla cars are really well connected to the internet, so they need to take care of security because they are likely to be targeted more than other cars.”

It’s worth mentioning that Synacktiv crushed the competition and took home the title of “Masters of Pwn” in the three-day event. The pro pen testers managed to eclipse second-place finisher Star Labs, 53 to 19.5, winning over half a million of the $1,035,000 prize pool, plus the Tesla Model 3.

OpenAI’s ChatGPT is a powerful and popular chatbot that can do all sorts of things, but it’s not the only example of its kind. Here are some alternatives you might want to try instead.

Bing Chat

Microsoft introduced chatbot-style search to the Bing search engine, which is being referred to as “the new Bing.” At the time of writing, you’ll need to get on the waitlist to use the feature, which you can do by signing up at Bing.com/new . In our experience, it only took around 24 hours to get approved. Once you’re in, the AI component of Bing is free to use.

At launch, you’ll need to download Microsoft Edge for desktop or mobile (or try our workaround for using Bing AI in Chrome ) and log in with your approved Microsoft Account . Finally, you’ll be able to ask Bing questions. A new “Chat” tab will appear at the top of the search results, while relevant queries will trigger a conversation to the right of your standard search results.

The new Bing works a lot like ChatGPT, answering questions as if you were having a real conversation. The search engine uses third-party sources like those presented in standard Bing search to inform its answers, which are listed alongside its claims. If the source is inaccurate, Bing’s answers to your questions will be too. It’s subject to the same pitfalls as any other AI chatbot , but links to sources mean greater transparency when compared with ChatGPT. Still, there are some undeniably cool things you can do with Bing AI .

Microsoft stresses that you should “use your own judgment and double check the facts before making decisions or taking action based on Bing’s responses” just as you would when searching the web. Bing can summarize information, list ideas, create plans, come up with written content (stories, poetry, advertisements, and so on), and even come up with code.

RELATED: 6 Cool Things You Can Do With Bing Chat AI

Google Bard

Google Bard is now available, too. Google Bard is Google’s answer to ChatGPT and Bing’s AI chat. You can sign up to use Google Bard on Google’s website . Unlike ChatGPT, Bard is essentially “hooked up to the internet” and is taking in new information all the time. That’s great for timely data, unfortunately, Google Bard seems less polished and capable than ChatGPT and Bing’s AI chat feature in its initial release.

RELATED: Google Bard: How to Use Google’s AI Chatbot

OpenAI GPT-3 Playground

Before ChatGPT, there was GPT-3 Playground. Powered by the same underlying models at ChatGPT, Playground provides a more customizable toolset that allows you to fine-tune the OpenAI experience. At the top of the screen, you’ll see some preset examples of how the tool works with the ability to save or share your conversations.

The tool has three modes: a prompt-based “Complete” mode, an “Insert” mode which can be used to insert text between two points using an [insert] tag, and an “Edit” mode which you can use to instruct Playground to fix text for grammar and spelling or debug code.

There are a few other options you can play with, including a “Temperature” slider to adjust randomness, a length slider for given responses, and more. You can use these settings to choose “cheaper” operations which cost less in terms of processing power. You’ll be needing those since Playground isn’t free to use forever.

OpenAI starts you off with $18.00 worth of credit, which eventually expires so you might as well use it. Each operation has a token cost, where 1000 tokens represent around 750 words. OpenAI’s Davinci models (which is what ChatGPT uses) have a running cost of $0.02 per 1000 tokens.

GitHub Copilot

ChatGPT is not only capable of conversing in plain English, but it can also be used to write code. AI advancements have not (yet) matured to a point where anyone can use them to write code regardless of knowledge, but tools like GitHub Copilot can be a huge help to aspiring developers.

Available with a 60-day free trial (and only $10/month thereafter), GitHub Copilot uses the same OpenAI Codex system as ChatGPT to assist programmers as they work. The tool was trained on “billions of lines of public code” to act as an assistant that can help a developer remain focused on the problem they’re trying to solve or the system they’re trying to build rather than the nitty gritty of programming.

GitHub is keen to state that Copilot is far from being a replacement for good coding or security practices. The tool can anticipate what a developer may want to achieve next based on supplied code or even natural language comments within the text.

The tool works with dozens of languages and can be implemented in development environments like Visual Studio Code, Visual Studio, JetBrains IDE, and Neovim. Verified students , teachers , and maintainers of popular open-source projects on GitHub can get access to Copilot free of charge .

Jasper

Jasper describes itself as an “AI Content Platform” with a focus on creating content to be posted on the internet. This includes everything from social media posts, articles, emails, website copy, classified listings, long-form content, meta descriptions, and more. Jasper can even be used to create AI art .

The platform also includes Jasper Chat, a chat interface that’s not dissimilar to ChatGPT. Unlike ChatGPT, Jasper isn’t free to use. The most you can hope for is a demo that gives you 10,000 words for free, and you’ll need to provide payment details to get started.

Play Video

Jasper is aimed at businesses looking for a tool that can generate copy and even comes with a 45-minute bootcamp that trains users on how to best use the platform. The developers are keen to stress that Jasper can pass most plagiarism tests, but any kind of AI content regardless should be checked for plagiarism and then proofread and fact-checked.

YouChat

Not to be confused with the instant messaging app of the same name, YouChat is a ChatGPT-like AI search assistant that can present information in a natural language conversation. It’s very similar to what Microsoft has done with the new Bing, except there are fewer hoops to jump through before you can use it.

The search engine chatbot is live right now. Simply head to You.com and ask a question, then click on the “Chat” tab. To have a back-and-forth conversation on this screen you’ll need to create an account and be logged in.

YouChat backs up many of its claims with links to sources, which are often top search results from websites like Wikipedia, product pages, or blogs. This is useful in that you can at least see where the information has come from, though it makes many unsupported statements.

YouChat is arguably faster than Bing and ChatGPT, but it also told us that “YouChat AI is powered by ChatGPT, an advanced language model developed by You.com” so your mileage may vary .

WriteSonic

WriteSonic is a lot like Jasper in that it is focused largely on creating and manipulating content. This includes writing articles, creating summaries, expanding upon existing content, creating descriptions, writing adverts for classified websites or products, website copy, and idea generation.

Unlike Jasper, you can get started with Writesonic for free without the need to provide any payment information. The website gives you 25 free generations via its ChatGPT-like “ChatSonic” chatbot every day, with up to 6,250 words included at no cost as part of the main Writesonic utility.

The tool is subject to the same potential pitfalls as any other AI content generator. Search engines like Google reward high-quality content that is written for “people-first” rather than purely to rank in search results. Expert fact-checking and input remains critical; these should not be treated as “set it and forget it” tools that can do everything for you.

Regardless of the business aspect, if you’re only looking for a ChatGPT-like chatbot that you can use for free every day then ChatSonic might be worth a shot.

Replika

Replika is an AI-powered companion that’s available on the web, on your smartphone, or via the Oculus VR virtual reality platform. Unlike ChatGPT, which is largely designed for learning, authorship, and discovery, Replika is aimed primarily at providing companionship.

Replika has been creating quite a stir online for its promise of providing an AI companion to those in need. The service raises all sorts of ethical questions about a chatbot that gives the illusion of getting to know you as a person. Replika asks you what your day was like, how you’re feeling, what sort of mood you’re in, and many of the other things you’d expect to hear from a concerned friend.

There’s something difficult about opening up to a chatbot that is very interested in your emotional state. The service assures you that data safety is not of concern and that your conversations are private, but you’d be forgiven for finding it difficult to tell Replika about your financial problems, medical diagnosis, or divorce proceedings without worrying about the myriad of ways that uploading all your problems to the internet could come back to haunt you.

Replika is the real-life equivalent of what we have seen many times over in (both dystopian and utopian) sci-fi novels and films, right down to the $19.99 monthly upgrade to Replika PRO to access certain conversation topics or upgrade your Replika’s status to “Romantic Partner” ( yes, really ).

Frequently Asked Questions

Is ChatGPT Free to Use?

ChatGPT is free for anyone to use right now. As the service grows in popularity, servers may experience high loads and so the service might not be available at all times. You can upgrade to ChatGPT Plus for $20 per month using the “Upgrade” button in the sidebar which gives you priority access to the chatbot.

Which AI Is Better Than ChatGPT?

AI-powered chatbots like ChatGPT are constantly evolving, so there’s no single answer to whether one is better than another. While ChatGPT is a powerful and interesting tool, Bing’s transparency makes it easier to fact-check its claims. Bing’s implementation (at the time of writing) uses OpenAI’s GPT-4 model, while ChatGPT is still using GPT-3.5. This makes Bing the more advanced chatbot, but that might not necessarily be true once ChatGPT gets upgraded too.

Is There a Good Open Source ChatGPT Alternative?

PaLM + RLHF by Philip Wang is an open-source alternative to ChatGPT that is (currently) untrained. It’s a barebones model that (to quote the project’s FAQ page) needs “millions of dollars of compute + data to sail to the correct point in high dimensional parameter space.” Downloading and running the model on your own machine is still a way off.

Can You Use ChatGPT Offline, or Is There an Offline Alternative?

Despite what the name may suggest, OpenAI is not an open-source foundation and thus there’s no ability to download ChatGPT or the (current) models on which it runs. You can’t run ChatGPT locally on your own machine, which means you can’t access ChatGPT offline either.

That’s not to say there aren’t models you can download and run locally, but you’ll need a fair bit of technical knowledge to use them. You can download pre-trained models like the open access BLOOM (which needs around 700GB of space) and then learn how to run it locally on your computer . You can download all manner of models from Hugging Face but getting them working is far from a user-friendly process.

If you want to learn, though, we’ve written a step-by-step guide to running a ChatGPT-like AI bot called Alpaca on your PC .

More ChatGPT Alternatives are on the Way

Google has already rolled out an initial version of an AI-powered chatbot called Bard , so soon the two biggest search engines will each have a stable ChatGPT alternative you can use. These AI-powered tools can do all sorts of things, from writing a cover letter to planning an event .