Why is the release of GPT-4 a big deal?

Abhinav Jain
3 min readMar 14, 2023

It’s actually NOT the pictured chart. Open-AI CEO Sam Altman called it “complete bullshit.”

Training on more parameters doesn’t necessarily make LLMs better. That’s a form of data ‘maximalism’ only 14% of AI experts believe. But there is one huge reason to be excited. Microsoft’s Germany CTO Andreas Braun spilled the news at a recent AI event.

“We will introduce GPT-4 next week, there we will have multimodal models that will offer completely different possibilities — for example, videos”

That’s a huge unlock for the system:

1. Images

Image generation has traditionally been the realm of tools like Dall-E, the first generative AI tool hyped last year. This represents GPT’s progress to becoming an “everything” AI.

It’s going to be much more than Dall-E.

Based on a paper Microsoft published, GPT-4 is expected to be able to READ images. So you will be able to ask it to OCR an image, or decipher the attitude of a painted face. Think of the power of a transform prompt like “age this person 10 years.”

2. Audio

It’s expected that GPT-4 will be able to have voice interaction as another modality. It will be quite fun to compare to Alexa and Siri. I expect it will fare very well. And unlike them, you’ll be able to ask it to write a song or transcribe a podcast.

3. Video

As Andreas mentioned, GPT-4 will be able to read videos as input and build them as output. One could imagine an entire show being made in GPT-4. Or YouTubers quickly editing their videos. Tons of visual & audio jobs are going to be affected by GPT-4. It will not just be the writers anymore.

The really crazy thought is: what about GPT-5? Or 10? GPT-3.5 has already passed the Turing test, Lovelace test… you name it, in our AI criteria. Who knows how powerful the next versions will be.

SARs & Phantom Stock: The New Age ESOPs for Consultants, Advisors and everyone in between. There are restrictions under law with respect to who can be issued ESOPs in the company, and being an employee of the company is a mandatory pre-condition.

But in a lot of startups, especially in the initial years — freelancers and external consultants may be adding a lot of value to fill in for unfilled co-founder roles such as Fractional CTO, Fraction CFO and so forth. If any advisor equity is to be granted to such people, it is mostly structured in the form of Stock Appreciation Rights or Phantom Stock.

“Phantom” because quite like a ghost, there is no real stock given to the advisor. The advisor’s payout though is linked to the valuation of the company — and this payout may be in the form of cash or equity — on a future date when certain milestones may be hit.

The stages are similar to those of ESOPs — grant, vest, exercise, sale…

One of the biggest benefit of SARs is that a company does not have to dilute any shareholding on the cap table for advisors, while giving them a share in the appreciation of value in the company in the form of higher cash settlement in the future.

The fact that such cash payouts may also be convertible to equity shares in the future, helps one manage cash flows as well, and makes this one of the most flexible structures to design compensation among consultants who add great value in your company.

Additionally, such SARs can also be used for issuance to employees who you do not want to grant the right to acquire actual shares in your company. We’ve drafted these in multiple Advisor Equity Agreements, but would love to know of any oddball edge case where you’ve seen this happen :)

Whatever you’re facing, lift up your eyes to the vision, not the problem.

--

--