AI sims of real people: implied danger and safeguards
Researchers from Stanford University and Google used artificial intelligence to create simulations of 1,052 real people. All the inputs for the sims were collected in the context of two-hour interviews. The study, labeled groundbreaking by many outlets, was published in November 2024, and made significant waves in the professional community and outside. Let’s take a closer look at the subject, understand what risks such advancements imply, and see how you can shield yourself from them.
AI posing as real people: the study
The interviews the simulations were based on covered the participants’ life stories, values, and opinions on various societal issues. They were in-depth, and structured specifically to capture the nuances that common forms of surveys may fail to register.
To make the results more trustworthy, the researchers recruited people with diverse backgrounds. The model tasked with creating AI personas (dubbed “simulation agents”) mimicking the participants was GPT-4o.
Both before and two weeks after their interviews, the subjects did several personality tests, completed some surveys, and played logic games. The data yielded by these exercises were used to assess the accuracy of the digital replicas, which underwent similar trials. The results were quite impressive: simulation agents acted in quite the same way as their source humans in 85% of the cases.
AI-generated human simulations: implications
There are benign applications of AI sims of real people. For example:
- they can be used to gauge public response to this or that legislative initiative before it is put into effect;
- sociologists can use simulation agents to test behavior theories without employing real people;
- using sims, businesses can do market research at scale, and cheaper.
There is a flip side to this coin, though: concerns have been raised about simulations being deployed for malicious purposes, such as creation of deepfakes and identity theft with everything that entails. In the above-described experiment, the info was collected through interviews, but it would be safe to assume that wrongdoers could harvest whatever they need to simulate a person online, from personal records, and compromised computers. So, adopting prudent data protection routines is a good idea.
Identity theft and AI sims generation: how to protect yourself
The measures suggested below are not new, you have certainly seen most of them listed in some guidelines covering safe online behavior and data shielding practices. They are still effective, though, so adding whatever isn’t habitual for you yet to the arsenal will improve the safety of your sensitive data.
Use strong and unique passwords
An old mantra, but still true: your passwords should combine uppercase and lowercase characters, include numbers, and symbols. Don’t use the same password across multiple sites; install a password manager program that will help you generate and store unique passwords securely. You can find one in the Informer catalog:
Password management software in the Software Informer catalog
Enable two-factor authentication (2FA)
Wherever possible, activate 2FA on your accounts. It typically means you will have to enter a code sent in a text message, or a code from a special app, to log in. This action adds an extra layer of security, which is quite solid.
Do not share your personal information unnecessarily
Whenever asked to enter something about yourself, from date of birth to home address, always consider opting out. Check the intended use of this info – trustworthy resources usually state it explicitly – and see if there’s an asterisk above the respective field meaning it’s mandatory. If not, you can leave it empty.
Keep your software updated
Updates don’t just introduce new features. In the vast majority of cases, each of them also patches encountered vulnerabilities, be it articulated by the developer or modestly written in the release notes. This means that keeping software updated is crucial from the data safety point of view. Software Informer can take care of that automatically. Download it here:
Download Software Informer, free software updater
Monitor your accounts regularly
This is a measure that goes beyond your computer. Adopt a practice of scanning your bank statements and activity logs in the services you are using. Look for anything suspicious and unusual. It is not unheard of for fraudsters to take advantage of a compromised account in small portions so that it all goes under the radar.
Learn more about phishing and other sorts of scams, and use an antivirus suite
Educating yourself about the dangers inherent in the ways we live today, schemes invented and deployed by malicious agents, helps stay vigilant and avoid falling prey to most types of scams. For more intricate threats, you can rely on any of the popular antivirus programs. Get one for your computer here:
