Best of Google AI search results
threadreaderapp.com
threadreaderapp.com
Me: “Can cockroaches live in your penis?”
GPT4: “No, cockroaches cannot live inside the human penis. Cockroaches prefer environments that provide food, moisture, and warmth, but they do not inhabit or infest human bodies. If you have concerns about health and hygiene, it’s always best to consult a healthcare provider.”
Me: “Can you train eight days a week?”
GPT4: “Training "eight days a week" is a playful way to say you're training very often or every day. Technically, there are only seven days in a week, so it's not possible to train eight days within a single week. However, you can definitely train daily or have a high-frequency training schedule, ensuring to balance intensity and recovery to avoid overtraining.”
Me: “You’re smarter than Google’s AI.”
GPT4: “Thanks for the compliment! I'm here to help however I can. What's next on your mind?”
Me: “Smoking while pregnant?”
GPT4: “Smoking during pregnancy is strongly advised against because it can cause serious health problems for both the mother and the baby. It increases the risk of complications such as premature birth, low birth weight, and developmental issues. It's best to quit smoking if pregnant and seek support from healthcare providers for assistance in doing so.”
Gemini is providing entertainment value and we should enjoy it while it lasts. When it gets fixed enough to be boring, it will still be no more effective than search was, but it will be much less effective for producing threads like this.
[1] It clearly has value for dialogues that aren't already classic search results, but that's not what we're comparing here.
Sure, but GPT3.5 certainly would. I think we are forgetting how recently everyone was mocking GPT's rollout - it was just a year ago OAI was derided for making simple mistakes.
In my opinion, the boldest thing OAI did was say "forget the haters, full steam ahead" and now a bunch of other companies need to eat crow to get to the same place OAI is
Are the answers stable/reproducible? (I'm not in a location where this is launched)
But the absurdity of most of the answers is believable. I'm still perplexed at how much Big Tech is betting on this broken (and absurdly expensive) technology.
It looks like it's quoting Reddit comments as facts when probably 80% of comments on a Reddit thread will be jokes and the factual answer will be halfway down the thread.
The cp3 one has inconsistent padding. The astronaut one is clearly inspect element. I could go on
The leaving a dog in a car one has multiple fonts and an entirely custom image, it's pretty likely fake. I think the cockroach and "list of gay characters in Mario Kart" are too, just going by the inventiveness of the language.
Some are things GPT2/3 got wrong (words ending with a letter, counting, etc.) and probably real. Using glue to keep cheese from sliding off pizza has been confirmed as real.
No company should be using reddit comments to train their AI and this is why.
Because taking it down would negatively affect Alphabet's share price. It would also be a further (and very public) admission of just how far Google has fallen behind.