Grok is a crock
Beware of what you believe
Whatever you know or think you know about AI , beware, they aren’t all the same. Spell check on a document or offering you TV shows on your streaming platforms is safe AI. Recently certain systems, including Grok , have fuelled violence and risk to life and limb. The old journalists adage used to be, a story is only valid if it comes from 2 independent sources. The journalists can then decide what to write, what to say is true and what is claimed to be true.
AI and chatbots have more sources but not always more reliable ones, also they will still try and answer questions without enough information. They want to please, they want to appear helpful. They are basically a slightly advanced search engine, its what they search that matters. Grok is different to most AI systems in that it employs real time access to find information. This live information is usually things shared on the X platform. This means that when there is enough fake posts on a subject, Grok is then influenced by false information.
This false information gets shared and then reposted, it can then strengthen Groks answers even more. Search engines look through thousands of sources, you may get a few dodgy sites in your answer but most are ok. Also with multiple sites offered then you can look for common answers and weed out what looks wrong.
In its eagerness to please Grok has been putting lives at risk recently. Following the Henry Novak murder, certain people wanted to know the police officers involved. Using photos these people asked Grok and Grok got it wrong. Not that mattered once it had given its answers. Protesters 1then forced 2 innocent people into hiding , although they weren’t involved in the incident. In fact , one of the people targeted left the force over a year before the incident.
Following the Renee Goode murder earlier this year, protesters again went looking for the ICE agents involved. Using scant information and various photos Grok got it wrong again. Another individual was targeted, although not involved and not an ICE agent. Again the information gets shared and acted upon, before anyone tracks it for voracity.
One of the many problems for chatbot and AI systems is that their positive points and be used for negative use as well. There was controversy earlier this year about the pornography the site created to order. Pictures of anyone could be uploaded and any style pictures could be created. Even child images were made, in all forms of sexual imagery. Celebrities, ex partners and random strangers could share the indignity, with the faked photos being shared for all to see.
AI presently has no human conscious, no thought of outcomes and no thought on dangers. Any time someone non-white commits a crime in the UK certain people demand to know the perpetrators ethnicity, immigration status and name. Often the information is wrong or at best flawed. People like Nigel Farage and Tommy Robinson revel in causing division and therefore share the information. When the terrible murder of 3 young girls happened at a dance class Farage called for banning immigrants. The murderer wasn’t an immigrant and was born in the UK.
As these people fan the flames of racism , the information is shared far and wide. As x has a habit of limiting posts that follow Musk’s beliefs , Grok then scrapes even more lies. The information then becomes self fulfilling, with certain people believing everything they read and the agitators that speak the same rubbish. When you only get information from one source , you don’t get the whole story.
Every story in substack where I need to check information, I look in multiple places to confirm information. I also limit myself to certain sites and try and avoid the dodgier looking ones. Unlike many trolls I wouldn’t use one site as my backup to anything I said or claimed. Unfortunately many of the trolls that I argue with will discredit however many sites I show links too, while deciding their single source is correct.
While AI can be useful for certain things, its not a safe place for general information yet. Remember when GROK turned off its safety checks and started to say how wonderful Hitler was. The systems need to be open to as much information as possible in order to get nearer the truth.
The main search engines don’t search when you ask them too. They have collated the information from thousands of sources and gets your answer from there. They look for questions and then check they have the information to answer. A search program will then look for all information based around the subject. For example when Renee Goode was murdered there wouldn’t have been much information to hand. Therefore when people asked about her the systems predicted questions that would be asked and found the information it might need.
With sites like Grok just looking round the present on a single source the information would be thin at best. With Musk interfering in foreign politics and sharing dangerous stories Grok becomes even more unsafe. Musk is no more than a Bond villain, albeit one with a wide reach.
AI is great at checking x-rays, good at telling you what to watch, its not good with unbiased and factual information.