If you’d like to read this issue on my website, click here! If you’d like to sign-up, and receive this in your inbox each week, click here! Read past issues here.
Good Friday Morning! Except the family of a woman on vacation in Hawaii. The NYPost tells us that she got selected for an upgrade on a flight to Hawaii, and took it. Whereas her brother, on the same flight and with a lower status with the airline, did not. It’s caused a split in the family, forcing this lady to stay with a friend, away from the family. It’s the definition of first-world problems.
There is obviously, one major bombshell of a news story on Thursday: Trump’s conviction in the NYC hush money case, guilty on all 34 counts. When the jury said it had a verdict this early, I was somewhat surprised at the speed. And that speed did not bode well for Trump. However, because I’m on vacation, I’m not covering that story. I’ll wait until I’m back stateside next week with better internet service to dive into things. I will give you my first takeaways, in no order:
- Trump is unlikely to see jail and has the makings of a successful appeal.
- Conservative’s relationship with the law just became much more interesting.
- Trump is likely to set a record in fundraising.
- The push to get a guilty verdict on Hunter became much more essential.
- Republicans will be sorely tempted to return fire and prosecute Joe Biden and Hilary Clinton. I don’t know what would stop them.
- Trump is likely to dip in the polls for a while, potentially giving Biden his first lead in months. I don’t believe that will hold. If Trump’s polling is unaffected by this story, Biden is in much deeper trouble electorally than I’ve suggested.
More next week, and I’ll flesh out these points more. For this week, I’m going into AI, and how bias is being shaped as it develops. Also, how AI will impact the kinds of news people will consume in the future – links to follow.
Quick Hits:
- The divide in the Democratic Party is getting starker. On the one hand, you have people like Rashida Tlaib, who are making surprise guest appearances at conferences with links to both the CCP and Palestinian terrorist organizations, calling Biden a supporter of genocide and worse. On the other hand, Senator Elizabeth Warren removed herself from a conference led by a group that has praised Hamas’ actions on October 7. These aren’t hard connections to make. The moves and speeches are public, for all to see.
Where you can find me this week
Please subscribe, rate, and review my podcast on iTunes, Spotify, or Google Play — the reviews help listeners, and readers like you find me in the algorithms. Make sure to sign up for the Conservative Institute’s daily newsletter.
We Need To Investigate Joe Biden’s Involvement In Hunter’s Trial – Conservative Institute
Biden Launches Campaign Hail Mary At Trump Trial – Conservative Institute
The Alito Flag-Gate Controversy Implodes In Left’s Face – Conservative Institute
Growing AI With Less And Less Data
I’ve written a lot on what AI is being used to do, and how it’s getting deployed in new and unique ways. A lot of this is bleeding-edge stuff that is hitting the mainstream in unexpected ways. What I’d to focus on now is one of the growing weak spots of AI, and how that is going to play into the potential for political and other biases down the road.
Put simply: we’ve known for a while that Google’s Search and the algorithms on social media are biased. These companies are trying to maximize profit with their products and also pushing the beliefs/morality of the company owners. There’s nothing new about those statements, and the critique is as old as mass media. There was a viral video on TikTok/Instagram/X(Twitter) this week where a girl talked about a viral post in her feed. When she looked at the comments, everything she was shown was stuff bashing men. When she gave the same post to her boyfriend, he was shown a radically different set of top comments. Nothing is ever as it seems.
This critique wasn’t always true of the internet. At the outset, it was largely an unregulated place where people could find or see whatever they wanted. While that version of the internet was closing by 2016, Trump’s election slammed it shut. The tech world and liberal media at large were traumatized by what “free speech run amok” had given the world. They’ve tried to cram the genie in the bottle ever since, and the fight continues.
Artificial Intelligence is currently in the Wild West phase. I know that it doesn’t seem that way given the various forms of bias people are finding in it, but it’s the truth. That’s slowly starting to change, however. And part of the problem is how AI models are trained.
To begin, to create a Large Language Model (LLM) like ChatGPT, you need to feed it data. Up until recently, all training for every AI model out there was done using open-source information. We have models trained on Wikipedia, non-copywrite works, and anything open-source on the internet. There are pros and cons to this, but to explain here is how some training is done:
We need a lot of data to train powerful, accurate and high-quality AI algorithms. For instance, ChatGPT was trained on 570 gigabytes of text data, or about 300 billion words.
Similarly, the stable diffusion algorithm (which is behind many AI image-generating apps such as DALL-E, Lensa and Midjourney) was trained on the LIAON-5B dataset comprising of 5.8 billion image-text pairs. If an algorithm is trained on an insufficient amount of data, it will produce inaccurate or low-quality outputs.
The quality of the training data is also important. Low-quality data such as social media posts or blurry photographs are easy to source, but aren’t sufficient to train high-performing AI models.
Text taken from social media platforms might be biased or prejudiced, or may include disinformation or illegal content which could be replicated by the model. For example, when Microsoft tried to train its AI bot using Twitter content, it learned to produce racist and misogynistic outputs.
This is why AI developers seek out high-quality content such as text from books, online articles, scientific papers, Wikipedia, and certain filtered web content. The Google Assistant was trained on 11,000 romance novels taken from self-publishing site Smashwords to make it more conversational.
To be useful, AI has to be accurate and current. There are growing questions regarding whether AI scraping data from news websites and training on it constitutes a copyright violation. I have no idea how the courts will settle that question, nor does anyone else.
But the open-source material out there is easy to find. However, it’s also finite. The current generation of AI models is growing restricted by the data they have access to. And, more concerning for these companies: less data means less ability to grow their AI capabilities.
That means these companies can either pivot to using low-quality data (social media sites, etc.) or get access to the paywalled content. OpenAI, the maker of ChatGPT, is aiming to get access to the paywall content. Here’s a summary of recent deals:
Two more media companies have signed licensing agreements with OpenAI, allowing their content to be used to train its AI models and be shared inside of ChatGPT. The Atlantic and Vox Media — The Verge’s parent company — both announced deals with OpenAI on Wednesday.
OpenAI has been quickly signing partnerships across the media world as it seeks to license training data and avoid copyright lawsuits. It’s recently reached deals with News Corp (The Wall Street Journal, the New York Post, and The Daily Telegraph), Axel Springer (Business Insider and Politico), DotDash Meredith (People, Better Homes & Gardens, Investopedia, Food & Wine, and InStyle), the Financial Times, and The Associated Press.
For OpenAI, the upside is obvious: they get access to all the current and past content to train their AI, which gives it a much greater repository to train on. In return, these companies will have their content featured as a source in ChatGPT when people are asking questions.
If you’re following along, you’ll note that these are all industry-leading names. A piece in Semafor notes that this AI push looks like a “revenge of the establishment media.” Instead of finding something via Google search, which will include any website, AI is limited to what is is given. If it doesn’t have access to other websites, it won’t train on or recommend them. Semafor writes:
The partisan publishers that thrived during the social media age — most of all, high-flying right-wing outlets like Breitbart News and the Daily Caller — are being shut out of the new AI boom.
But while some on the left groused that the News Corp. package includes the right-leaning tabloid New York Post, the true impact of the new marriage of AI and news appears to be the revenge of the establishment media. And fringier, more explicitly ideological outlets on the right have noticed that their businesses — already rocked by an industry-wide decline in web traffic — seem unlikely to get an AI bailout.
“It does concern me, but not surprise me, that these left wing tech companies are ignoring news outlets trusted by half of America and focusing exclusively on training AI models using left leaning news sources,” said Neil Patel, Daily Caller co-founder and publisher of the Daily Caller News Foundation, in an email. “The result will of course be shockingly biased AI systems having influence and even control over many aspects of life. The insanity that Google released is just the tip of the iceberg. It’s flat out scary.”
The Daily Caller isn’t alone. The Daily Wire, founded by the conservative pundit Ben Shapiro, hasn’t fielded any inquiries either, an executive there said.
They’re not just leaving out conservative news outlets. BuzzFeed has been working with OpenAI to develop content for its site using generative AI. But one person at the company told Semafor that’s where the relationship ends. The AI company has not attempted to do a content licensing deal for HuffPost, the prolific left-leaning news site owned by BuzzFeed.
The internet utterly destroyed the traditional media model. When people were given a choice on how to get their news and daily information, the media world shattered into a million pieces. The fact that you’re reading this newsletter here and not some mainstream site is proof of that reality.
As an aside, the New York Times is headed in the opposite direction of this flow. They’re suing OpenAI over copyright infringement. The Times is a big enough brand that they can afford to be the oddball out, but I suspect we’ll end up seeing them settle this case, and cut a deal, too. The Times won’t want to get left out of the next digital revolution, especially if it means losing its top status in media.
What AI is presenting legacy media is the chance to surge back into the lead, and stiff-arm the new competition. They can claim these partisan outlets are “low-quality” information, and that AI shouldn’t get trained on it. In turn, these outlets get to be the premier outlets again, shutting the door on competition.
The next point is obvious: news bias would get worse under this system, not better. The moment competition and criticism get removed from the equation, the news sites training AI will dramatically impact not only the information people read, but also how AI interprets the data it’s being fed. Right now, we’re dealing with the morality issues of tech-bro culture in California. And although it’s biased, it’s also extraordinarily stupid. When AI is asked to create a picture of the Nazis, and can’t do it without making all depictions of black people and women in that garb, you’re dealing with a deeply unserious and dumb bias.
The bias is news media is different because bias isn’t just in how these outlets report something, but what they don’t report. The benefit of having conservative media is that they can report on things the large outlets leave out because it’s “conservative news,” or “inconvenient” for liberal ideology. If these outlets don’t report on something that happened, it will get frozen out of AI models.
Ultimately, the question will be: what kind of bias do we end up with artificial intelligence? Avoiding bias is impossible. There’s going to be something. I’d rather it reflect a “Wild West” approach, like the early internet. But I also know that’s unlikely. The internet opened up information and granted access to everyone. We are bizarrely watching things close back down as people start viewing the world through the lens of the primary AI model they use.
As we move along, AI is a tool everyone will use in their jobs. And it’s impossible to know how we’ll settle all these things. But that’s the state of things. Ask me how this works in six months to a year from now, and I might give a different answer. It’s all building quickly, so we won’t have to wait long.
Links of the week
Trump trial precedent: Will it be open season on candidates? – NewsNation
Hunter Biden’s Chinese Client Appealed to CCP for Help in Legal Case, Court Records Show – Washington Free Beacon
Israel evicts UNRWA from country – YNet
San Francisco quietly removes ‘Appeal to Heaven’ flag from outside City Hall after Alito flap – Fox News
Lock ’em up libs run the Democratic Party – Timothy P. Carney, Washington Examiner
I just donated $300k to Trump. I’m prepared to lose friends. Here’s why. – Shaun Maguire
X/Twitter Thread(s) of the week
Satire of the week
Trump Boys Attempt To Bribe Juror With Briefcase Full Of Grape Uncrustables – Onion
Retired Angel Hernandez Gets New Job Inspecting Planes For Boeing – Babylon Bee
Woman Checking Email Will Be Disappointed if There Is None, Disappointed if There Is Some – Reductress
Pope Francis Offered Three Netflix Stand-Up Specials Following Use of Gay Slur – The Hard Times
John Fetterman Staff Confirm Senator Has Successfully Respeced Entire Character – The Hard Drive
Dad Enters ‘Sure It’s Always My Fault’ Phase Of Life – Waterford Whispers News
HR Ask Worker Sobbing In Office Bathroom To Keep The Noise Down – Waterford Whispers News
Thanks for reading!