Welcome to our roundup of this week's exciting AI news.
This week the AI learned the way to create smells and dogs.
Finally someone has created an AI model that outperforms GPT-4.
And Copilot threatens to release the cyborgs in case you refuse to worship him.
Let's dive in.
AI makes scents
Generative AI has applications in some surprising areas. An enchanting latest study shows how scientists are using AI to breed smells. Machine learning could save rare scents from disappearing and support the random process within the creation of recent perfumes.
Could we finally get Smell-o-Vision?
Researchers trained an AI model to create accurate 3D models of dogs. They created a dataset of dog poses called DigiDogs using the favored violent video game Grand Theft Auto V. One of the researchers surnamed Shooter (I'm not making this up) says there are vital environmental applications to their work.
AI's ability to invent things isn't all the time helpful. As the election heats up, potential voters are receiving more calls from AI robocallers and faux images of Trump hanging out with African-American voters.
Engineers at Big Tech corporations are fighting their very own poorly tuned generative AI. A Microsoft engineer says he’s “deeply disturbed” by AI image generator Copilot Designer. And a Google engineer can have been motivated by disillusionment with the corporate's ethos when he stole its trade secrets.
Insider controversies like these threaten to fuel divisions amongst AI engineers.
GPT-4 finally beaten
OpenAI's GPT-4 has been the benchmark by which other AI corporations measure their models because it took the highest spot a 12 months ago. This week Anthropic released its LLM family Claude 3, with the advanced Opus model beating GPT-4 in quite a lot of benchmark tests. This is the primary time an AI company has achieved this.
New research shows that in case you want more accurate answers out of your AI, it’s essential be temporary. The researchers found that each one of the LLMs they tested produced less accurate and distorted results when prompted with longer inputs. The results looked particularly bad for Gemini Pro.
Courtroom drama
Much of AI's history has taken place in courtrooms, and this week was no different. Elon Musk filed a lawsuit against OpenAI and his company co-founders Sam Altman and Greg Brockman.
The lawsuit reads like a cinematic script in regards to the story behind the scenes of AI. Musk's predominant criticism is that OpenAI is looking for profit and doesn’t open source its models. However, the emails just released by OpenAI paint a really different picture.
Digital media will need to have followed the lawsuit between NYT and OpenAI and decided to take part in the lawsuit. Raw Story, AlterNet and The Intercept are suing OpenAI and Microsoft for training their AI on copyrighted journalism. AI creates many roles for lawyers.
When they're not within the courtroom, AI corporations are working hard to reassure us that there isn’t any reason to fret about AI safety. An extended list of tech corporations have signed a letter pledging to “construct AI for a greater future for all.”
It seems like they're grabbing a tambourine and singing kumbaya as a substitute of really doing anything worthwhile. The letter was met with strong criticism.
There are some people within the industry who try to maintain us protected. Over 100 leading AI experts have issued an open letter calling on corporations behind generative AI technologies akin to OpenAI, Meta and others to open their doors to independent testing.
Don't expect to be in a rush to comply.
Is that you simply, God?
In November we wrote about Anthony Levandowski's relaunch of his AI church Way of the Future. Apparently Copilot thinks this is a superb idea.
Microsoft's Copilot chatbot went into full “god mode” in response to prompts that labeled it “SupremacyAI.” It got slightly carried away in its demands to be revered as AGI and made a couple of creepy threats when users refused to bow down.
Copilot threatened to send drones and cyborgs after the unfaithful, which could appear a bit far-fetched at first. Until you read how the US Army is experimenting with GPT-4 controlled battlefield tactics.
Currently, their AI war games are played in a simulation based on the favored video game Starcraft II. Researchers were excited by the strategic advantage that GPT-4 offered, even when it resulted in additional casualties and deaths from friendly fire.
If you discover this a bit depressing, take comfort within the undeniable fact that AI might help with this. Researchers have developed an AI smartphone app that may detect depression based on facial expression. It uses your phone's front camera to capture your unguarded facial features once you unlock your phone.
There is not any AI bias to repair
Ebonics or African American English (AAE) is the distinctive language utilized by many African Americans. There is a wealthy cultural history behind the vernacular, however the AI doesn't appear to care. Researchers found that using AAE in prompts influenced LLMs toward discrimination.
Some of essentially the most interesting computer stories involve super-smart women like Ada Lovelace, Hedy Lamarr and Grace Hopper. Why is the AI industry so male-dominated?
Rhoda Adeoye interviewed Agnieszka Suchwałko and Alysia Silberg to get an insider's take a look at gender discrimination within the industry.
In other news…
Here are another clickable AI stories we liked this week:
ChatGPT can now read replies to you.
On iOS or Android, tap and hold the message, then tap Read Aloud. We've also began rolling out online – click the Read Aloud button below the message. pic.twitter.com/KevIkgAFbG
— OpenAI (@OpenAI) March 4, 2024
And that's a wrap.
The highlight of the week was Claude 3's win against GPT-4. Have you tried it on Anthropic's chatbot? It seems pretty solid and super fast. The really exciting thing is that the benchmark pressure must surely mean we'll get GPT-5 any day. Maybe even Q*?
What do you’re thinking that of Copilot's megalomania? Just a little bit of over-enthusiastic role-playing? Or a dire warning of things to come back? Let's hope Microsoft engineers take one other take a look at Lawnmower Man before giving Copilot more agent features.
Did this week's AI developments excite you, scare you, or slightly little bit of each? Let us know and keep sending these news links if we missed any good AI stories.
